New Hosts of The Lassa Virus

Lassa virus (LASV) causes a deadly haemorrhagic fever in humans, killing several thousand people in West Africa annually. For 40 years, the Natal multimammate rat, Mastomys natalensis, has been assumed to be the sole host of LASV. We found evidence that LASV is also hosted by other rodent species: the African wood mouse Hylomyscus pamfi in Nigeria, and the Guinea multimammate mouse Mastomys erythroleucus in both Nigeria and Guinea. Virus strains from these animals were isolated in the BSL-4 laboratory and fully sequenced. Phylogenetic analyses of viral genes coding for glycoprotein, nucleoprotein, polymerase and matrix protein show that Lassa strains detected in M. erythroleucus belong to lineages III and IV. The strain from H. pamfi clusters close to lineage I (for S gene) and between II & III (for L gene). Discovery of new rodent hosts has implications for LASV evolution and its spread into new areas within West Africa.


Results and Discussion
Lassa virus in Hylomyscus pamfi, Nigeria. In October 2008 and March 2009, a preliminary screening for arenaviruses was conducted during the process of describing the African wood mouse H. pamfi as a new species 10 . Of 10 specimens captured in Kako, southwestern Nigeria, during this period, 4 were LASV positive (Table 1). Partial GP sequences of these strains, labeled Nig09-OSPMH86, 89, 123 and 124, have been used to develop a LASV-specific diagnostic test 11 . At that time, only limited amounts of tissue were available, preventing the isolation of the virus. As these sequences clustered with Lassa lineage I (Lily Pinneo strain) in the partial L phylogeny (Supplementary Figure 2), we decided to go back to the field to catch new animals aiming to obtain a positive specimen that would allow us to perform complete sequencing of the genome. Fortunately, in March 2012, one H. pamfi out of 2 captured in Kako was PCR-positive for LASV (Table 1), and the virus was successfully isolated in our BSL4 laboratory. Throughout our sampling in Kako, no M. natalensis was PCR-positive.
Phylogenetic analysis of nucleotides of the complete GP and NP genes revealed that the virus strain from Kako clustered close to lineage I (the Lily Pinneo strain), while in phylogenies based on complete L and Z genes the Kako strain clustered between lineages II and III (Fig. 1). Bayesian phylogenies, maximum likelihood and neighbor-joining trees (data not shown) exhibited similar topologies for all genes with high bootstrap support values at most branches. Although we detected no evidence of recombination or reassortment events among various segments of the LASV lineages, phylogenetic analysis based on the amino acid sequences of the L and Z genes clustered the Kako strain differently in a sister relationship but still distant to the singleton Pinneo strain of lineage I or with the members of the lineage II ( Supplementary Fig. 1). Collectively, our phylogenetic results (with the Kako strain falling among LASV lineages I-III but not clustering exclusively to any of them) indicate that Kako might constitute a potential lineage of its own. Additional sequences obtained from rodent specimens and possibly also from humans will help to establish this new lineage.
The amino acid differences between the Kako strain and the singleton Lily Pinneo strain is 13.3% for the NP segment, higher than the 12% cut-off suggested by Bowen et al. 12 as a delimitation criterion between LASV and other arenaviruses 12 . However, as increasingly diverse Lassa strains are recovered from humans and rodents (e.g. 7,8,13,14 ) the 12% cut-off for designating strains belonging to LASV will most likely have to be revised. In fact, a recent study that generated complete LASV genomes from Nigeria and Sierra Leone 14 reported high nucleotide sequence variation of up to 32% and 25% for the L and S segments respectively, higher than the findings in Bowen et al. 12 .
In addition, detection of the Kako strain supports the view that LASV is more diverse than previously thought and is certainly a species complex, with strains emerging that fall within the lineages I-IV but are hosted by rodents other than M. natalensis. From an evolutionary perspective, Kako represents the link between LASV and other closely related arenaviruses such as Mobala virus and Mopeia virus discovered in M. natalensis in eastern Africa 15,16 , and also the more recently detected Gbagroube virus and Jirandogo virus both found in Mus spp. in western Africa 17,18 .  Table 1). All 3 virus strains were isolated in the BSL-4 lab and sequenced for the whole genome. Phylogenetic analyses of the nucleotide sequences from complete GP, NP and polymerase genes show a clustering with Lassa CSF within lineage III, with highly supported nodes (posterior probability = 1, Fig. 1). The same result was obtained at the amino acid level (Supplementary Figure 2), with divergence between Onmba Abena and Lassa CSF of 5.8-6.2% for the NP segment.

Lassa virus in
In Nigeria LASV lineage III, as detected in humans, circulates in the north central area of the country 20 . The Lassa CSF strain, which falls within lineage III, was also detected in a patient in this area, on the Jos Plateau 21 . As we found CSF-like LASV in M. erythroleucus in Onmba Abena, which lies to the immediate south of the Benue River and close to the distribution range of LASV lineage III detected in humans, this suggests that M. erythroleucus could possibly also be a host of LASV lineage III in north central Nigeria. Again

Lassa virus in Mastomys erythroleucus, Guinea.
In 1991-92, a spatial survey on LASV seroprevalence in humans showed a high rate in Madina Oula, which is located in the coastal area of Guinea (35%, 59/171) 22 . Various small mammal surveys performed through 1983-2009 show that coastal Guinea is outside the geographic range of M. natalensis 23,24 . The human and murine data have been compiled and mapped in Supplementary  Figure 3. This discrepancy between high human seroprevalence and the possible absence of the reservoir invited us to verify on the field this unusual epidemiological scenario. In May 2014 we captured sixteen M. erythroleucus in Madina Oula, while not a single M. natalensis was caught. Six animals were LASV PCR-positive (prevalence 37.5%), and phylogenetic analysis of the nucleotide sequences from complete GP, NP, polymerase and Z segments of these BSL-4-isolated strains show that they belong to lineage IV (Fig. 1). In all trees, the Madina Oula sequences clustered to those from Sierra Leone (Josiah, NL and LM395) and forest Guinea (Z148 and Z158) rather than to those from Upper Guinea (Bantou). Phylogenetic analysis of the amino acid sequences shows the same clustering for 3 proteins, except for the GP (Supplementary Figure 1).
Arenavirus antibodies have been detected very recently in M. erythroleucus in three other localities within coastal Guinea, with two of these villages being geographically close to Madina Oula in the same prefecture:   M. natalensis and M. erythroleucus are closely related species, with sympatric populations across West Africa (supplementary Figure 3) 25 . Their role as co-reservoirs of LASV indicates host-switching via recurring spill-over infections, as has been increasingly demonstrated for other virus-rodent models by various authors 26,27 . It has been shown that repeated spill-over infections through contact with a donor species can lead to adaptation of a virus and its emergence in an alternate, recipient host 27 . This scenario is supported by serological study mentioned earlier in Guinea 23 in which Lassa antibodies were detected in M. erythroleucus in localities where, on one hand, LASV-positive M. natalensis were present (presenting the possibility of spill-over infections) and also in other localities within coastal Guinea, on the other hand, where M. natalensis was completely absent.

Conclusion
Our discovery of alternate rodent reservoirs for LASV in H. pamfi and M. erythroleucus is quite surprising, considering previous large scale investigations in a country like Guinea that appeared to convincingly dispel this notion 6  This study provides increased insight into the evolution of LASV and demonstrates that the virus is more complex genetically and ecologically, maintained by multiple reservoirs. Our investigations have implications for the epidemiology and control of Lassa fever. As the ecology of M natalensis (which is mostly commensal) is not exactly the same as that of H. pamfi (forest dwelling) and M. erythroleucus (more of a generalist) the potential is present for Lassa fever to emerge in fresh niches other than currently exist in West Africa.

Methods
As part of surveys screening small mammals for Lassa virus, rodents were trapped in Kako, south western Nigeria (N07° 41′ E04 37′ ); Onmba Abena, eastern Nigeria (N7° 38′ E8° 24′ ); and Madina Oula, coastal Guinea along the border with Sierra Leone (N9° 52′ W12° 26′ ) (Fig. 1). Ranging from October 2008 to May 2014, the sampling dates for each locality are detailed in Table 1. Each date represents a session of 3 trapping nights per locality using Sherman live-capture traps. In Nigeria permission to trap rodents in Kako was granted by the Osun State Ministry of Environment and in Onmba Abena by the Gwer West Local government Area, Benue state. In Guinea permission to trap rodents was obtained from the national ethic committee (12/CNERS/12). The methods were carried out in accordance with the approved guidelines. Total RNA was extracted from whole blood (Nigeria) or dried blood on filter paper (Guinea) employing a QIAamp viral RNA Mini Kit (Qiagen Inc.). Extracted RNA was tested for Lassa virus by RT-PCR to amplify the GPC and L genes 11,28 . Cytochrome b gene sequencing of all Lassa virus-positive rodents unambiguously identified them as Hylomyscus pamfi and Mastomys erythroleucus. Additionally, all Mastomys from Nigeria were sequenced for cytochrome b to exactly distinguish specimens of M. natalensis from M. erythroleucus. Lassa strains from PCR-positive samples were cultured in the BSL-4 laboratory at the Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany. The whole genome for each of the Nigerian strains was obtained by Next Generation Sequencing (NGS), and those from Guinea via Sanger technology.
Complete genome sequencing of Lassa virus strains. The full-length genomic sequence of Nigerian LASV isolates was determined using next-generation sequencing technology. The cell-culture supernatants were filtered through a 0.45-μ m filter (Millipore, Darmstadt, Germany) to remove larger debris and bacteria and digested with a mixture of nucleases (Turbo DNase, Ambion, Carlsbad, CA, USA; Baseline-ZERO, Epicenter, Madison, WI, USA; Benzonase, Novagen, San Diego, CA, USA; RNAse One, Promega, Fitchburg, WI, USA) to digest unprotected nucleic acids including host DNA/RNA. Enriched viral particles were then extracted, fragmented, reverse-transcribed, ends repaired, dA-tailed, adaptor ligated and purified. Library preparation was performed using NEBNext ® Ultra ™ DNA Library Prep Kit for Illumina ® (New England Biolabs, Inc. USA).
The Sanger sequencing of the Guinean LASV strains complete genomes has been performed as follows: RT-PCRs were performed in a thermocycler Seqlab Primus 25 by using the OneStep RT-PCR kit (Qiagen) with a battery of different primers (Supplementary Table 1). Volume reaction was 20 μ l: 9 μ l water, 4 μ l one step buffer, 0.8 μ l dNTP, 1.2 μ l Fwd primer, 1.2 μ l Rev primer, 0.8 μ l enzyme and 3 μ l RNA. The cycling conditions were: 30 min at 50 °C, 15 min at 95 °C, 45 cycles including 30 sec at 95 °C, 30 sec at 52-55 °C and 1 min at 72 °C. The loop for S segment was obtained with a nested PCR by using Superscript III OneStep RT-PCR (Invitrogen) and Platinum Taq (Invitrogen). Specific outer primers were designed to amplify this fragment: LVSmad 1208+ , and LVSmad 1985-(Supplementary Table 1). In round 1, volume reaction was 25 μ l: 3.5 μ l water, 12.5 μ l one step buffer, 1.  Table 1).
Scientific RepoRts | 6:25280 | DOI: 10.1038/srep25280 Phylogenetic analysis. The phylogenetic analyses were performed using Bayesian Markov chain Monte Carlo tree-sampling methods based on 2 runs consisting of 4 chains of 1,000,000 with a burn-in of 25% using MrBayes v3.1.2 (http://mrbayes.sourceforge.net/) and parallel maximum likelihood with PhyML v3.0 (http:// www.atgc-montpellier.fr/phyml/versions.php) and Neighbor-Joining methods with 1,000 pseudo-replicates. The Akaike information criterion was chosen as the model selection framework and the general time-reversible model of sequence evolution with gamma-distributed rate variation among sites and a proportion of invariable sites (GTR+ Γ + I) for nucleotide sequences and Johnes-Taylor-Thorton with gamma-distributed rate variation among sites and a proportion of invariable sites (JTT+ Γ + I) for amino acid sequences as the best model. All model selection methods employed, jModelTest 2 33 , Topali v2.5 34 and MEGA6 35 detected the same model of sequence evolution that fit best the data sets. All LASV sequences were confirmed as non-recombinant by using the various methods for recombination detection implemented in RDP3 36 and Phi test in SplitsTree v4.12.3 37 .
All virus and rodent sequences generated by the study have been submitted to GenBank with accession numbers KT992416-KT992450, and KM052324-KM052326.