PCR products of the expected size were amplified from 11 of 40 HPS patient samples (Table 1) and 7 of 25 rodent samples studied (Table 2). A 303-nt fragment of the G2 gene was amplified and sequenced (bases corresponding to position 2807 to 3109 of Lechiguanas [LEC] virus). Phylogenetic analysis of nucleotide sequence differences showed the cocirculation of two genetic hantavirus lineages previously characterized from humans only: ARA virus and a genotype compatible with the previously identified JUQ virus. ARA virus sequences were derived from eight samples: from three HPS patients and three rodents (Bolomys lasiurus) from the state of São Paulo, and 1 HPS case-patient and one rodent (B. lasiurus) from Minas Gerais. Pairwise comparisons of the sequences of ARA virus strains from HPS patients and B. lasiurus showed an 85.1%-99.7% nt and 95%-99% amino acid (aa) identity. The viral sequence from the HPS patient Hu237251 from Patrocínio, Minas Gerais, the northernmost location included in this study, was most divergent from the other members of this group (16.7%). The second hantavirus genetic lineage identified was closely related to JUQ virus. Samples from seven HPS patients and three Oligoryzomys nigripes from the states of Rio Grande do Sul, Santa Catarina, Paraná, and São Paulo fall into this group (Figure 2).

Sequence comparison with JUQ virus, limited to an overlapping piece of 139 nt of the G2 encoding region, the only available from this virus (19), showed 85.7%-97.7% nt identity (Figure 3). The maximum identity with the prototype JUQ virus was found to correspond to the viral RNA from the patient Hu239727 (97.7%) from Barra do Turvo, São Paulo, the closest location to the site of infection of the fatal HPS case from which JUQ virus was originally characterized in Juquitiba in 1993. Sequences in JUQ-like clade group in two subclades, one including JUQ prototype strain and Hu239727, and the other one comprising six human and the three O. nigripes sequences (distance 0.273). The distance observed between these two subclades is intermediate between in-group and between group distances (Table 3).

Comparison of the 303-nt G2-encoding region sequences derived from HPS-patient and O. nigripes samples of this group showed 86.6%-99.7 % nt identity. Sequence comparison with the corresponding to ARA group showed 20.8% nt and 4.9% amino acid (aa) divergence. Genetic distances between the sequences studied are shown in Table 3. The general time reversible model (GTR) with a 0.2479 proportion of invariable sites and γ = 0.4121, was used in analysis. AND virus resulted in ARA closest sequence (mean distance = 0.516) and Hu39694, JUQ-like closest sequence (0.723). Mean distances between ARA and JUQ-like sequences further support that they are different viruses.

A longer fragment of 1,239 nt of the G2 encoding region of the M segment (Figure 4), as well as a fragment of 259 nt of the nucleoprotein encoding region of the S segment, was generated by one representative strain of each hantavirus group (Bl194307 for ARA virus and On193576 for JUQ-like virus). Comparison of viral RNA from B. lasiurus sequences with ARA virus showed a 96.8%-nt and 99.3%-aa identity for the 1,239-nt G2-M piece, and 98.8% nt and 100 % aa identity for the S segment piece. When RNA viral sequences from B. lasiurus were compared with those from O. nigripes, 77.8% nt and 93.4% aa identity for the 1,239-nt G2-M piece, and 84.1% nt and 98.8% aa identity for the S segment piece were observed.

The phylogenetic relationship of ARA and JUQ-like genotypes to other hantaviruses of South America was determined on the nucleotide sequences of the 303-nt and 1,239-nt G2-encoding region of the M segment genomes by using maximum parsimony and neighbor-joining analysis. Both trees showed a similar topology, except for the altered placement of certain genotypes that displayed low bootstrap support. Phylogenetic analysis performed on the 303-nt sequence fragment (Figure 2) showed that all samples from B. lasiurus fell into the ARA virus group, whereas all samples from O. nigripes fell into the JUQ-like genetic group. The nodes separating these groups had high bootstraps values (72% and 91%, respectively); however, the exact branching order among the Brazilian and the other hantaviruses cannot be resolved by the present phylogenetic analysis. The phylogenetic tree, based on the analysis of the 1,239-nt M segment sequence differences, showed the same topology of the 303-nt sequence, data-based tree. Both maximum parsimony (Figure 2) and neighbor-joining (data not shown) analysis demonstrated that the divergent ARA and JUQ-like viruses form a unique monophyletic clade with the other South American hantaviruses (71% bootstrap support).

Within this clade the ARA virus forms a subclade with other two akodontine-borne Argentinean hantaviruses, Maciel (MAC) from Necromys benefactus and Pergamino (PRG) from Akodon azarae, although with a low bootstrap support (66%). The JUQ-like virus from O. nigripes 193576 is ambiguously placed within the South American viruses, together with CAS virus from Brazil and the rest of the Oligoryzomys-borne Argentinean hantaviruses (LEC, Orán, Hu39694, Bermejo, AND virus), as well as LN virus from Paraguay. A slight difference in the topology of this 1,239-nt database tree was observed when neighbor-joining analysis was used; the grouping of this Oligoryzomys-borne Argentinean hantavirus clustered in a subclade with a bootstrap support of 79% (data not shown).

Eleven (27.5%) of 40 human samples and 7 (28.0%) of 25 rodent samples studied tested positive for hantavirus by RT-PCR. Previous studies have characterized three different hantavirus genetic lineages associated with HPS cases in Brazil: CAS, ARA, and JUQ (20,30,31).

Previous serologic studies detected IgG antibodies in rodent biologic samples: in B. lasiurus and Akodon spp. captured in the state of São Paulo (32), as well as B. lasiurus from São Paulo and Minas Gerais States, and O. nigripes from São Paulo, Rio Grande do Sul, Santa Catarina, and Paraná States (33).

The data we report on the phylogenetic analysis of viral M and S genome segment fragments, from HPS patients and rodent samples from different locations of southern and southeastern Brazil, showed the circulation of two distinct hantaviruses, closely related to ARA and JUQ viruses, previously characterized only from humans. Our data on the phylogenetic analysis of a 303-nt G2 encoding region of the M genome segment represents the first genetic evidence of the role of B. lasiurus as rodent host reservoir for ARA virus, as well as O. nigripes as rodent host reservoir for the JUQ-like virus in the region under study. The nodes separating each group of virus within the South American hantaviruses clade were highly supported (72% and 91% bootstrap for ARA and JUQ-like lineages, respectively). The two subclades observed in Figure 3 (i.e., 139-nt tree), and the intermediate distance between them suggested the possibility of some geographic isolation. More extensive sequencing of JUQ prototype and JUQ-like sequences is needed to clarify this point.

Comparison of the sequences of ARA virus strains obtained from rodents (B. lasiurus) and HPS patient samples showed an identity of up to 99.7% at the nucleotide level, while within the JUQ-like virus group, the comparison between O. nigripes and HPS patient virus strains showed an identity of up to 97.7%. However, the phylogenetic relationship of JUQ-like hantavirus to other members of the South American hantavirus lineages, determined on the nucleotide sequences of the 303-nt sequence, as well as on a longer fragment of a 1,239-nt piece of the M segment performed for one representative strain derived from one O. nigripes, systematically failed to resolve the branching order.

In South America, human illnesses associated with hantaviruses have been linked to viruses from the Oryzomyini and Phyllotini tribes. Known akodontine-borne hantaviruses have not been associated with human illnesses. Thus, the data reported here on ARA rodent reservoir constitute the first evidence that a hantavirus associated with an akodontine rodent can cause HPS. Phylogenetic tree based on a 1,239-nt G2 sequence fragment places together the akodontine-borne ARA virus from B. lasiurus in Brazil, MAC from N. benefactus, and PRG from A. azarae in Argentina (34). Although the bootstrap support displayed was low (66%), this finding is in accordance with previous observations based on the S genome phylogeny of Argentinean hantaviruses (35). This would support the hypothesis of cospeciation of hantaviruses with their specific rodent hosts. As it has been described with other hantaviruses, biogeographic factors are also involved in the evolution of hantavirus lineages (36,37). The human- and rodent-derived ARA strains analyzed in the current study were distributed at a distance of approximately 650 km. As expected, ARA virus strains originated from the more distant localities displayed the highest genetic divergence, as shown between samples from Patrocínio (Hu237251) and Uberlândia (Bl235018) in Minas Gerais State, and Flórida Paulista (Hu196618) in São Paulo (16.7%-nt difference). Similarly, the divergent JUQ-like virus sample Hu239727 originated from Barra do Turvo, São Paulo, in relation to the rest of the human and rodent JUQ-like virus samples included in this group from locations in Santa Catarina, Paraná, and Rio Grande do Sul may be associated with the geographic distance between them (500 km on average).

Although the data from serologic testing by ELISA indicated four positive Akodon spp., one C. tener, and one Ox. rutilans, specific viral sequences could not be amplified from those specimens. Thus, additional studies are needed to determine the possible role of these species in the epidemiology of hantavirus in Brazil.

The habitats and behavior of the rodents are important aspects to consider in elucidating the reservoirs of etiologic agents. ARA virus was recovered mostly from HPS patients as well as B. lasiurus samples from the ecosystem called cerrado, while JUQ-like virus was recovered mostly from human and O. nigripes samples from the ecosystem called Atlantic rainforest. Geographic distribution of B. lasiurus in Brazil includes the original areas of cerrado and "caatinga." This environment is typical in northeastern Brazil, characterized by deciduous trees and cactus and an extremely prolonged dry season. The B. lasiurus distribution shows its ability to adapt to anthropic environments, especially grasses (Brachiaria) and sugar cane cultures. These rodents are aggressive and usually dominate the areas they infest (38); they do not colonize human dwellings, although occasionally they can invade houses.

O. nigripes is adapted to live in the primary and secondary forests, especially in the Atlantic rainforest and Araucaria forest. It is primarily found in anthropic environment, such as the lineal natural habitats bordering cultivated areas, especially those with corn, where it is the most abundant species. These rodents can easily invade dwelling houses and barns to search for food, and they can nest in the domestic habitats.

Among the rodents captured in the cerrado, B. lasiurus was the most abundant species (44%) and showed the highest prevalence of antibodies to hantavirus (11%). Akodon spp. and O. nigripes were the two most abundant among those rodents captured in the transition area between cerrado and the Atlantic rainforest, but the highest prevalence of antibodies to hantavirus was found in O. nigripes specimens (8%) (33). These data help incriminate B. lasiurus and O. nigripes in the transmission of ARA and JUQ-like viruses, respectively, to humans.