From October 13 through December 3, 2008, a total of 555 small mammals were trapped in Morogoro, Tanzania (6.84°S, 37.65°E). This period corresponds to the end of the dry season, when the density of M. natalensis mice is usually high (7). Sherman traps were set in habitats where these mice were expected to occur in high density (7). Dried blood samples were preserved on calibrated, prepunched filter papers. Blood samples (1 punch ≈15µL/rodent) were eluted in 300 µL of phosphate-buffered saline and tested for antibodies to arenaviruses by indirect immunofluorescence antibody (IFA) testing using Morogoro virus as antigen. In addition, total RNA was extracted from another punch of blood by using 300 µL of AVL buffer (QIAGEN, Venlo, the Netherlands). The lysate, plus 300 µL of ethanol, was centrifuged in a silica column (Zymo Research, Orange, CA, USA). The column was washed with 400 µL of AW1 and AW2 buffers (QIAGEN). RNA was eluted with 15 µL of water. A 1-step reverse transcription–PCR (RT-PCR) selective for the Morogoro virus RNA polymerase (large [L]) gene was performed as described (6); it was based on a pan–Old World arenavirus RT-PCR approach (8), but primers were adjusted to detect the Morogoro virus (6). Products were shown by agarose gel electrophoresis. A subset of the amplicons was purified and unidirectionally sequenced by using MoroL3359-fwd primer (6). Amplicons derived from Lemniscomys rosalia and Mus minutoides mouse blood samples were bidirectionally sequenced. Nucleotide and amino acid sequences were aligned by using BioEdit software (9). Old World arenaviruses and 2 representatives of New World arenaviruses were used to estimate nucleotide and amino acid pairwise divergence (p-distance) with MEGA 4 (10). A phylogram was reconstructed by using the neighbor-joining algorithm in MEGA 4.

We trapped 511 M. natalensis mice and 44 individuals from 7 other small mammal species (Table 1). IFA results were positive for 58 blood samples from M. natalensis, 1 from L. rosalia, and 1 from M. minutoides mice (Table 1). M. rosalia mice were trapped in woodlands, whereas M. minutoides mice were trapped in vegetable gardens and fallow fields. These results are consistent with recently reported results from a study in the same locality 20 years ago, in which mice from the genera Lemniscomys and Mus were seropositive for arenaviruses according to IFA with Lassa virus as antigen (6). In our 2008 study, the antibody prevalence for M. natalensis mice was 12.1%, which is low compared with 50% antibody prevalence reported for 2004 and 2007 (6), suggesting high fluctuation of interannual or seasonal prevalence of Morogoro virus in its host.

Using RT-PCR selective for the arenavirus L gene, we obtained positive results for 43 mice: 41 M. natalensis, 1 L. rosalia, and 1 M. minutoides. In total, 6 samples were positive according to IFA and RT-PCR. We sequenced 33 RT-PCR amplicons. The sequences derived from M. natalensis mice (286 bp used for the analysis) showed 97.1%–100% amino acid homology with the Morogoro prototype L sequence (GenBank accession no. EU914104). In contrast, the 2 sequences derived from the blood samples of L. rosalia and M. minutoides mice showed only 69.3% and 65.2% aa homology with the Morogoro prototype L sequence. These sequences (320 bp) were compared with sequences of the Old World arenaviruses (Table 2). The virus amino acid sequence from M. minutoides mice clustered at 93.7% homology with that of the Kodoko virus, in the lymphocytic choriomeningitis clade (Figure; Table 2): thus, the arenavirus of M. minutoides mice seems to be a strain of the Kodoko virus originally isolated from 2 M. minutoides mice in Guinea (11). Our finding supports M. minutoides mice as the true reservoir of Kodoko virus in Africa.

The amino acid sequence of the virus isolated from L. rosalia mice clusters with the Ippy virus sequence (Figure). Ippy virus was isolated in the Central African Republic from Arvicanthis niloticus rodents (12). For the portion of L gene sequenced (320 bp), the level of amino acid divergence between the 2 is 17.3%, higher than the level of divergence between other Old World arenavirus species (e.g., 14.5% aa divergence between Mobala and Lassa viruses; Table 2). Thus, the arenavirus found in L. rosalia mice appears to be a new species of Old World arenavirus. The genus Lemniscomys is more closely related to the genus Arvicanthis than to the genera Mus and Mastomys.

In high-density habitats of M. natalensis mice, where Morogoro arenavirus transmission occurs, sympatric murine species do not seem to be secondary reservoirs for the virus. In contrast, 2 mouse species, L. rosalia and M. minutoides, seem to be reservoirs of 2 other Old World arenaviruses, 1 of which may be a new species. Our study emphasizes the complementary nature of serologic and genetic-based approaches for arenavirus detection. Because of the cross-reactivity of Morogoro antigens with immune serum from individuals infected with other arenaviruses, a serology-only approach might have led to the conclusion that an extended set of hosts exists for the Morogoro virus. Because of its high cost, a genetics-only approach might never have indicated the hot spot of arenavirus around Morogoro that was shown by IFA (6). However, critically, genetics then allow cross-reactivity to be decomposed.

Our study demonstrates the presence of 3 Old World arenaviruses in a single location. To date, only 5 Old World arenavirus species and 17 New World arenaviruses have been recognized by the International Committee for Taxonomy of Viruses (13). Although the likely presence of additional arenaviruses in Africa has long been suggested (14,15), the discovery of new Old World arenaviruses is rare. Our study illustrates that arenaviruses in Africa may be highly diverse and demonstrates the efficiency of the recently developed pan–Old World arenavirus RT-PCR for identifying new Old World arenaviruses (8). To isolate and describe the new arenavirus of L. rosalia mice and the strain of Kodoko virus, additional sampling and genotyping are being conducted. In particular, determining the sequence of the S segment will further clarify evolutionary relationships within the Old World group.