Ecology, Genetic Diversity, and Phylogeographic Structure of Andes Virus in Humans and Rodents in Chile

Human samples.All patients examined were officially enrolled in a study that was approved by the human institutional review boards of the University of New Mexico and Pontificia Universidad Católica de Chile. Patients with symptoms of HCPS were considered to have acute ANDV infection on the basis of the following serological criteria: the presence of immunoglobulin M antibodies against the viral N antigen, with or without specific immunoglobulin G antibodies. We identified patients through a national passive surveillance/diagnostic referral system administered through the Institute of Public Health in Santiago, Chile (18). We randomly selected patient blood or tissue samples for further study according to availability and geographic diversity.

Rodent samples.We collected rodents in live traps according to standard protocols, as previously described (63). We followed established safety guidelines for rodent capture and processing (39). Trapping grids were set up at sites adjacent to case households and in locations chosen to provide geographic diversity. Animals in the field were anesthetized using ketamine, blood was drawn from the retro-orbital sinus, and the rodents were then sacrificed via cervical dislocation. We necropsied rodents on-site and collected and preserved hearts, kidneys, spleens, livers, and lungs in liquid nitrogen. A subset of rodent tissues was obtained through a national survey overseen by the Chilean Institute of Public Health.

Rodent antibody screening.Rodent antibody against ANDV was detected in blood samples using a strip immunoblot assay as previously described (63, 66). We used positive serology to identify individuals for viral sequence determination.

RT-PCR.Heminested reverse transcription-PCR (RT-PCR) was conducted essentially as described previously (21). Typically, 50 to 100 mg of heart or lung tissue or 100 μl of serum was used to extract RNA using an RNeasy minikit or a QIAmp viral RNA minikit (Qiagen Inc., Valencia, CA), respectively. We optimized S- and M-segment primers (Table 1) according to ANDV sequences ascertained through GenBank. We used the SuperScript II (Invitrogen, Carlsbad, CA) enzyme and 2.5 μM of an antisense primer to reverse transcribe ∼1 μg of total RNA or 5 μl of RNA derived from serum. For the S segment, the antisense primer (coordinate S 626) was used to carry out the RT step, followed by an outer PCR using 5 μl of cDNA and 100 μM of the sense (coordinate S 35) and antisense S626 primers in a final volume of 50 μl. We used either a Taq gold polymerase (Applied Biosystems, Foster City, CA) or the Taq DNA polymerase (Roche, Indianapolis, IN), and reaction mixtures were run in a PE Biosystems model 9700 cycler under the following conditions: 5 min at 94°C; 35 cycles, with 1 cycle consisting of 10 s at 94°C, 10 s at 48°C, and 60 s at 72°C; followed by a final extension step (5 min at 72°C) and a soak step at 4°C. Five microliters of each outer product was then subjected to 94°C for 5 min and 35 cycles (with 1 cycle consisting of 10 s at 94°C, 30 s at 50°C, and 60 s at 72°C), with a final extension step (5 min at 72°C) and a soak step at 4°C using 100 μM of the inner heminested primers S184+ and S626− in a final volume of 50 μl. The products were agarose gel purified and directly sequenced in both directions. We increased the length of S-segment sequences (to 943 nucleotides [nt]) in a subset of O. longicaudatus samples (pruned data set, 14 individuals). These samples were selected as representative samples to cover the geographic range of ANDV-seropositive individuals in Chile (Table 2). We amplified and sequenced two portions of the M segment, referred to as Gn (490 nt; coordinates 26 to 516) and Gc (577 nt; coordinates 2543 to 3119). The Gn gene was amplified by a heminested RT-PCR, using primer Gn559− for the RT step as described above and primers Gn1+ and Gn559− for the outer reaction and primers Gn1+ and Gn540− for the inner reaction. PCR conditions were similar except that the annealing temperature and time in the outer reaction were set at 48°C and 30 s, respectively. We used a nested RT-PCR to amplify the M-Gc segment using primer Gc3171− for the RT step, and primers Gc2488+ and Gc3171− for the outer PCR. Primers Gc2517+ and Gc3143− were used for the inner reaction. The conditions for the outer reaction were as follows: 5 min at 94°C; 3 cycles, with 1 cycle consisting of 15 s at 94°C, 30 s at 45°C, and 60 s at 72°C; 33 cycles, with 1 cycle consisting of 15 s at 94°C, 30 s at 50°C, and 60 s at 72°C; followed by an extension step (5 min at 72°C) and a soak step at 4°C. The nested reaction was conducted with an initial step of 5 min at 94°C; 35 cycles, with 1 cycle consisting of 15 s at 94°C, 30 s at 52°C, and 60 s at 72°C; with a final extension step (5 min at 72°C) and a soak at 4°C. Samples were analyzed over a period of 6 years. To avoid experimental error, false-positive results, or cross-contamination, we handled only up to five samples at one time. Negative- and positive-control samples, as well as parallel amplification reactions wherein the reverse transcriptase enzyme was excluded in the reverse transcription step were included in each run. No sequences were found to match our positive-control sample sequence (RNA extracted from ANDV strain CHI-7913-infected Vero E6 cells). We removed the primer sites from every sequence to perform all analyses.

Recombination analysis.The occurrence of recombination events among ANDV sequences was evaluated using the Genetic Analysis for Recombination Detection Algorithm (GARD), a genetic algorithm with an automated phylogenetic detection of recombination tool as implemented in the HYPHY package (datamonkey.org) (33). The algorithm screens a multiple-sequence alignment for evidence of discordant phylogenetic signs, identifying putative recombination breakpoints and nonrecombinant regions. We screened for potential recombination of 38 S-segment sequences (397-nt) and 17 S-segment (943-nt), Gn (490-nt), and Gc (577-nt) segment sequences used for phylogenetic analyses. When potential recombination between sequences was detected, analyses were performed using the nonrecombinant regions independently.

Phylogenetic analyses of the S segment.We aligned and manually edited 397 nt of informative S-segment sequence from 38 samples (Table 2) using ClustalW and BioEdit v7.0.5, respectively (20). Maximum parsimony (MP) and maximum likelihood (ML) optimality criteria were used for phylogenetic reconstruction using PAUP* (59). For MP, we used a transition/transversion ratio (Ts/Tv) of 3:1. Heuristic searches were performed for both optimality criteria, with 100 (MP) and 10 (ML) random stepwise additions and tree bisection reconnection branch swapping. Node support was evaluated with 5,000 nonparametric bootstraps for MP and 1,000 nonparametric bootstraps for ML (17). For ML analysis, we used Modeltest 3.7 (47) to determine the model of nucleotide substitution that best fits the data. The Akaike information criterion was chosen as the model selection framework, and the TrN model plus invariant sites (I = 0.5171) and gamma distribution (0.9727) was selected as the best model (61). Bayesian analyses were performed using the Markov chain Monte Carlo sampling procedure as implemented in MrBayes 3.1.2, based on the selected nucleotide substitution model obtained for ML searches (26). Four independent runs of four Metropolis-coupled chains of 5 × 10⁶ generations each were performed to estimate the posterior probability (PP) distribution. Topologies were sampled every 1,000 generations to ensure the independence of successive trees. The first 500 trees of the sample were removed to avoid including trees sampled before convergence of the Markov chain, and the last 4,500 trees were used to compute a 50% majority rule consensus tree. The percentage of samples that recover any particular clade on this tree represents that clade's posterior probability; we considered a P of ≥95% as evidence for significant support (2). We used the median-joining method (5) to perform a phylogenetic network analysis with ANDV Sout alleles (in-group; Table 2), as implemented with Network 4.2.0.1 software (http://www.fluxus-engineering.com/sharenet.htm) to further assess intraspecific relationships.

Phylogenetic comparison of the M and S segments.We performed maximum likelihood, maximum parsimony, and Bayesian analyses for a pruned data set (Table 2) of the S and M (Gn and Gc) viral segments, using the same phylogenetic parameters described above. Analyses with Gn and Gc sequences were performed including all five Andes virus lineages previously reported (43). The Gc sequence of the sample from Curicó, Chile, was not included in phylogenetic analyses due to our inability to amplify this fragment. Our analyses of S-segment sequences included only four ANDV lineages with no sequence available for the ANDV Cent Plata lineage (Table 2). The models of sequence evolution that best fit the data for each analysis were as follow: for the S-segment partition 1 (1 to 333 nt), Hasegawa Kishino Yano model plus gamma (0.1331); for the S-segment partition 2 (343 to 942 nt), transversion model plus gamma (0.1559); for M-Gn sequences (490 nt), general time reversible (GTR) model plus gamma (0.7458) plus proportion of invariant sites (0.4); and for M-Gc sequences (577 nt), GTR model plus gamma (1.1634) plus proportion of invariant sites (0.5666). In MP analysis, for each segment, the transition/transversion ratio was as follows: Ts/Tv = 6 for the S-segment partitions 1 and 2, Ts/Tv = 4 for the M-Gn sequences, and Ts/Tv = 4 for the M-Gc sequences. We also performed ML, MP, and Bayesian inference analyses with the translated amino acid sequences (200 amino acids) for the N protein (partition 2). Maximum likelihood was performed with TreePuzzle software (55) using the Dayhoff plus amino acid frequencies (F) model as calculated by ProtTest (1). We performed a phylogenetic analysis with concatenated sequences of the pruned data set (n = 21) that included sequences of the S, M-Gn, and M-Gc (1,668-nt) segments. Given that potential recombination was detected in the S segment (see Results), we used S-segment partition 2 (600 nt) and all Gn and Gc sequences for the concatenated analysis. Missing data were added for Curicó, Chile, in the Gc sequence. To estimate the PP of the phylogenetic trees, we used the Bayesian Markov chain Monte Carlo sampling procedure using BayesPhylogenies (45), which allows the combination of multiple genes and models into one data set without a priori partitions. We used a general likelihood-based “mixture model” based on the GTR of gene sequence evolution to estimate the likelihood of each tree. We determined the likelihood of the trees by first using a simple GTR matrix (nQ) and then using nQ+Γ mixture models where n varied between one and six independent rate matrices (Qs) and Γ represents a gamma-distributed rate heterogeneity model. A 50% majority rule consensus tree was constructed.

Out-group choice.For phylogenetic analyses, we included multiple prototypical viruses containing sequences overlapping with those of the S-segment portion sequenced in this study (Table 2). Trees were rooted using the SNV-S sequence. For the pruned data set, we selected out-groups that had previously been reported to be closely related to the in-group (ANDV) that had available sequences of S and M segments in GenBank (Table 2).

Molecular diversity and demographic analyses.We evaluated the molecular variability of the grouped S-segment sequences (397 nt) in the two primary clades (Fig. 1) using DnaSp 4.0 (52). We assessed population equilibrium of the two clades by performing Tajima's D test (60), Fu's F_S test (19), and R₂ test (49), testing the significance of the statistics from 10,000 simulated samples. We also examined the population dynamics for the ANDV S segment using the Bayesian skyline plot (BSP) (16) as implemented in the BEAST program (15). This method is a nonparametric coalescence analysis and uses standard Markov chain Monte Carlo sampling procedures to estimate past population dynamics from the posterior distribution of the effective population size (N_et) (14), with no a priori assumption of a specific demographic model. N_et can be thought as the number of ANDV infections contributing to new infections. Analyses were performed comparing uniform rates across branches (strict clock) and uncorrelated relaxed clock assumptions and using the model of sequence evolution as derived from the Modeltest program. Analyses were performed using a fixed mean substitution rate of 2.84 × 10⁻³ recently reported for the S segment of hantaviruses (50), and the results were plotted together with the number of confirmed human cases as reported yearly by the Chilean Ministry of Health (http://epi.minsal.cl/).

• Open in new tab
• Download powerpoint

FIG. 1.

Phylogenetic relationship and diversity of ANDV S-segment alleles in Chile. (A) Map of the sampled localities (numbers from Table 2) of the Andes virus in Chile and in the East Andes in Argentina). Black circles represent samples sequenced in this study, and black squares are samples obtained from GenBank. (B) Maximum likelihood tree obtained from aligned sequences (397 nt) of the S segment. Numbers above branches indicate pseudoreplicate bootstrap values obtained by maximum likelihood analysis (the first value [outside parentheses]) and maximum parsimony analysis (value in parentheses). Numbers below branches indicate posterior probabilities for Bayesian inference. The bar indicates 0.1 nucleotide substitution/site. Taxa shown in italic type are representatives of previously reported ANDV lineages as described in Table 2 and in Results. For further details, refer to Materials and Methods.

We calculated the mean number of nonsynonymous (dN) to synonymous (dS) nucleotide substitutions within each clade by using the Nei-Gojobori method implemented in the PAML 3.15 program (65). Because only a few sites may be under selection and to compensate for any lack of statistical power of this approach for detecting selection (32), we also included three complementary approaches (single-likelihood ancestor counting, random effects likelihood, and fixed effects likelihood), which estimate the rates of nonsynonymous and synonymous changes at each site in a sequence for detecting sites under selection (32). In addition, the partitioning approach for robust inference of selection (53), a maximum likelihood method that takes into account recombination and synonymous rate variation, was used to infer whether a proportion of sites in the alignment evolve with a dN/dS ratio of >1. Analyses were performed with various levels of significance, and the methods are implemented in the Datamonkey web-based gateway (30), which executes the molecular evolution analysis platform HyPhy (31).

Nucleotide sequence accession numbers.The sequences from all the ANDV Sout S and M segments described in this study have been deposited in GenBank and have been assigned accession numbers EU241637 to EU241653 and EU241655 to EU241702 (for details, refer to Table 2). Other previously published sequences used in the study were obtained from GenBank (Table 2).

Rodent seroprevalence and study design.During the years from 2000 to 2006, 7,230 rodents and other small mammals were captured throughout Chile as part of an NIAID-funded International Collaboration in Infectious Diseases project. Tissue samples from 3,304 wild rodents were frozen. The sampling scheme involved much of the range occupied by Oligoryzomys longicaudatus from Carrizal Bajo in the northern desert (28°14′S) to as far south as Torres del Paine National Park in southern Patagonia (51°S) in Chile. Of the total number of specimens analyzed, 103 rodents were seropositive for ANDV by using a strip immunoblot assay, with an overall seroprevalence of 1.4% (Table 3). O. longicaudatus displayed the highest seroprevalence (5.9%), followed by Abrothrix longipilis (1.9%). In both species, males had significantly higher seroprevalence rates than females (9.9% versus 1.91%, respectively, for O. longicaudatus and 2.6% versus 0.92%, respectively, for A. longipilis). Other rodent species showed much lower seroprevalence (≤0.6%), with only one or two animals testing positive for ANDV (Table 3). Of all the seropositive animals, tissue samples were available for 56 rodents from which viral RNA isolation was attempted. We were able to amplify and obtain viral sequences from 23 individuals (Table 2). Fifteen additional specimens were also amplified and sequenced from human and rodent samples obtained through the Public Health Institute of Chile (Table 2). Sequences were obtained from four different rodent species: O. longicaudatus, Rattus rattus, Abrothrix longipilis, and Loxodontomys micropus.

ANDV in Chile consists of a single cosmopolitan lineage.To assess phylogenetic relationships of ANDV in Chile, we used 38 viral sequences from human and rodent samples comprising the entire geographic distribution of HCPS cases and throughout the natural habitat of the colilargo in Chile (Fig. 1A). We found no evidence of recombination within the S-segment (397-nt) data set. ML, MP, and Bayesian inferences recovered all viral alleles in a monophyletic group, highly supported by Bayesian inference (PP = 1.00), but weakly supported by maximum likelihood bootstrap (MLB = 52) and maximum parsimony bootstrap (MPB = 57) values (Fig. 1B). All Chilean samples and the western Argentinean sample (El Bolsón, allele TN) located near the Andes mountains, were included in the ANDV Sout clade (in-group). Hantavirus samples from central, northern, and eastern Argentina were placed as a basal group outside the ANDV Sout clade, with Lechiguanas (ANDV Cent Lec), Hu39694 (ANDV Cent Bs.As.), and Oran (ANDV Nort) samples grouped in a supported clade (PP = 1.00; Fig. 1B). Within the in-group, two groups largely corresponded to two major ecogeographic regions in Chile, the Chilean matorral (shrubland; Mediterranean clade) and the Valdivian temperate forest (Temperate Forest clade). The Temperate Forest clade was well supported (PP = 0.98), while there was weak support for the Mediterranean clade (Fig. 1B). Within the Temperate Forest clade, samples were further subdivided into southern and northern Temperate Forest clades. Human samples and those of the four rodent species were placed in trees according to their ecogeographic origin. Abrothrix longipilis alleles TL and TS and L. micropus (allele TT) were placed in the Temperate Forest clade, while R. rattus (allele MH) was placed in the Mediterranean clade. Human samples were placed in either the Mediterranean (alleles MF, MG, and MJ) or Temperate Forest (alleles TA and TB) clades.

The unrooted network resulting from median-joining analysis (Fig. 2) is concordant with the tree resulting from maximum likelihood analysis (Fig. 1B) showing two major groups (separated by 25 mutations), with alleles associated by their ecogeographic distribution in either Mediterranean or temperate forest. Within the temperate forest group, samples from the south and north were further separated by 13 mutations (Fig. 2). Together, these results suggest that at least two clades occur throughout the geographical distribution of ANDV Sout in Chile.

• Open in new tab
• Download powerpoint

FIG. 2.

Unrooted network of ANDV Sout S segment using 397 nt with alleles depicted according to geographical distribution. The size of the circles represents the number of individuals per allele, and the designations represent alleles (Table 2). White and gray symbols represent alleles from the Chilean Mediterranean and temperate forest ecoregions, respectively. The origins of the samples are indicated as follows: squares, humans; circles, O. longicaudatus; rhombus, A. longipilis; pentagon, L. micropus; triangle, R. rattus. The circle and rhombus indicate that an allele was shared between O. longicaudatus and A. longipilis.

Phylogeographic relationships of ANDV in Chile.To begin to test for the presence of reassortant viruses and to better establish the phylogeographic relationships of ANDV in Chile, we compared the phylogenies of the M-Gn, M-Gc, and S segments obtained from a subset of 14 colilargos (Fig. 3). For the S segment, we used 943-nt sequences to perform sequence identity analysis. Not surprisingly, Chilean ANDV Sout S-segment sequences showed a high level of similarity that ranged from 96.8 to 100% at the amino acid level and 84.9 to 100% at the nucleotide level (Table 4). Identity ranged from 91.4 to 100% (amino acids) and 83 to 99.7% (nucleotides) for Gn and from 97.3 to 100% (amino acids) and 85.5 to 100% (nucleotides) for Gc (data not shown). Analyses of the S segment for potential recombination events showed evidence of discordant phylogenetic signals and detected a putative recombination breakpoint between nucleotide positions 334 and 342. Hence, S-segment phylogenetic inferences were performed independently with bipartitioned sequences using nucleotides 1 to 333 (partition 1) and nucleotides 343 to 942 (partition 2) of our sequence alignment. To determine presumptive recombinant sequences, we conducted analyses that excluded one sequence in every run. The recombination signal disappeared when one of three sequences were excluded: Tucapel, Los Angeles, or Panguipulli (sample 104535) (all of these were components of the Temperate Forest clade). By contrast, no evidence of recombination was evident for the Gn and Gc sequences.

• Open in new tab
• Download powerpoint

FIG. 3.

Phylogenetic comparison of the S and M segments of ANDV Sout in Chile. Maximum likelihood phylogenetic trees from aligned sequences of the pruned data set are shown. (A) S-segment phylogenetic tree partition 1 (1 to 333 nt), (B) S-segment phylogenetic tree partition 2 (600 nt), (C) M-Gn phylogenetic analysis (490 nt), and (D) M-Gc phylogenetic tree obtained using 577 nt. The numbers next to the branches are the pseudoreplicate bootstrap values obtained by maximum likelihood (first value) and maximum parsimony (second value) analyses and posterior probabilities obtained with Bayesian inference (third value) (−, no value). Names in italic type are previously reported ANDV lineages (Table 2). Recombining sequences are underlined in panels A and B. The bars indicate 0.05 or 0.1 nucleotide substitution/site.

When rooted with Laguna Negra virus, all phylogenetic analyses recovered ANDV Sout as monophyletic and with high support (Fig. 3). Within the ANDV Sout clade, the topology obtained using S-segment partition 1 recovered Mediterranean samples basal to the monophyletic Temperate Forest clade (Fig. 3A); S-segment partition 2 yielded two reciprocally monophyletic clades with high support, corresponding to the Mediterranean clade (MLB = 88, MPB = 98, and PP = 1.99) and the Temperate Forest clade observed previously (MLB = 54, MPB = 98, and PP = 0.88; Fig. 3B). When using the amino acid sequence of the N protein, the clades (Mediterranean and Temperate Forest) were less well supported in phylogenetic inferences (data not shown). Phylogenies using the Gn (Fig. 3C) and Gc (Fig. 3D) nucleotide sequences showed the same pattern by recovering two main groups within the ANDV Sout clade. The Mediterranean clade was supported using the Gn (MLB = 100, MPB = 99, and PP = 1.00) and Gc (MLB = 88, MPB = 95, and PP = 1.00) sequences. The Temperate Forest clade was supported by the Gc (MLB = 65, MPB = 65, and PP = 0.96) and Gn sequences using maximum parsimony (MPB = 99) and Bayesian inference (PP = 0.99).

Bayesian inference of the concatenated sequences of Gn, Gc, and S portions (1,668 nt) recovered both the Temperate Forest and Mediterranean clades with high posterior probability values (PP = 1.00; Fig. 4). Within the Mediterranean clade, two supported groups were recovered, representing northern (PP = 1.0) and southern (PP = 0.96) Mediterranean subgroups. Samples from the northern temperate forest ecoregion are derived, while the Chile Chico sample is basal (but poorly supported) to all samples of this clade (Fig. 4). Overall, these data confirm that at least two primary clades of ANDV Sout exist throughout the geographical range of this virus in Chile.

• Open in new tab
• Download powerpoint

FIG. 4.

Phylogenetic relationship of concatenated S- and M-segment sequences of ANDV Sout in Chile. Majority rule (50%) consensus tree for the 2Q+Γ model of concatenated M- and S-segment sequences (1,668 nt) using Bayesian inference. The consensus tree is based on 4,500 samples from a converged Markov chain. The numbers above branches are the posterior probability values. Sequence names shown in italic type indicate previously reported ANDV lineages (Table 2). The bar indicates 0.1 nucleotide substitution/site.

Molecular diversity and demographic analyses of ANDV S segment in Chile.Estimations of allele and gene diversity and results of Tajima's D test, Fu's F_S test, and R₂ tests are summarized in Table 5. Analyses were performed by grouping S-segment samples into “Mediterranean” and “temperate forest” groups. Overall, Fu's F_S test, Tajima's D test, and R₂ test were nonsignificant for both groups, as well as for the complete data set, indicating that the hypothesis of stable demographic history cannot be rejected in favor of a recent expansion. Nucleotide diversity was similar among both groups, with 0.056 (±0.003) for the Mediterranean clade and 0.048 (±0.004) for the Temperate Forest clades (Table 5). BSP analysis (Fig. 5) revealed, however, that the effective number of infections through time increased exponentially from 1999, reaching a peak during the years 2001 to 2002, after which the growth rate decreased. When the total number of human HCPS cases per year are superimposed on the BSP (data from the website http://epi.minsal.cl/), we observed a proportional relationship among the number of HCPS cases reported and the number of effective ANDV infections through time, with a ∼1-year delay observed in the latter (Fig. 5).

• Open in new tab
• Download powerpoint

FIG. 5.

Number of ANDV infections estimated from genetic data. The Bayesian skyline plot is derived from S-segment (397-nt) sequences. The thick solid line is the median effective number of ANDV infections (left-hand y axis) and represents the product of effective population size and generation time (in years). The shaded area indicates the 95% highest posterior density (HPD) region interval. Bars represent the total number of human HCPS cases per year (1997 to 2004; right-hand y axis) as reported by the Chilean Ministry of Health.

Within ANDV Sout, the dN/dS ratios for the S-segment sequences using PAML were ≪1.000 in the two clades, with a mean dN/dS ratio of 0.003 in the Mediterranean and temperate forest sequences (Table 5). These results were congruent with three complementary approaches (single-likelihood ancestor counting, random effects likelihood, and fixed effects likelihood) and the partitioning approach for robust inference of selection method (Table 5), which showed no evidence for positive selection, indicating that most substitutions were silent, and likely reflect strong purifying selection acting over the S-segment gene.