Diversity and Distribution of Parasite Lineages

One of the first steps for understanding the evolution and dynamics of host-parasite communities is to characterize their diversity and structure [11]. In the present study, we performed the first molecular characterization of the malarial parasites in bird communities from southeast Brazil. The overall prevalence (35.3%) observed was similar to other community-level studies that used PCR screening approaches and were performed in tropical habitats of South America [16]–[17], [26]. However, the observed prevalence was higher than that found in previous studies that adopted only blood smear screening approaches [27]–[31]. Furthermore, considerable differences were observed between our prevalence data and the results reported by Sebaio et al. [30], which also sampled birds from southeast Brazil. We attribute these differences to the intrinsic efficacy variation of the screening methods, which has been observed in other studies [16], [32]–[33].

In addition, we found that Plasmodium prevalence was higher than Haemoproteus prevalence. Although this finding complies with the previous study performed in southeast Brazil, most of the studies in bird communities of South and North America show a higher prevalence of Haemoproteus [11], [17], [31], [34]–[36]. This contrary scenario presented by the malarial parasite community from southeast Brazil could be explained by differences of vector populations, Culicidae mosquitoes for Plasmodium and biting midges (Culicoides) and louse-flies (Psedolynchia, Microlynchia and Ornithomyia) for Haemoproteus [2], because the distribution of malaria vectors can be affected by vegetation type, rainfall patterns, mean temperatures, elevation, and geomorphology [37].

Sequencing the mtDNA cyt-b marker revealed a greater number of lineages compared to other regional surveys conducted in South America, Europe and Africa [26], [36], [38]–[40]; however, the population had similar diversity compared to those from Australo-Papuan region and India [10], [41]. The high lineage richness discovered in this study could be associated with the diverse population of host species in southeast Brazil because parasites were found in 132 different bird species. These findings suggest that the high diversity of hosts, in conjunction with the heterogeneity of environments, is a source of niches in which malarial parasites can specialize and diverge as distinct evolutionary lineages.

The variable host range observed for both Plasmodium and Haemoproteus species suggests that their lineages have different strategies of host exploration, ranging from complete generalism to high levels of host specificity. Alternatively, considering that each vector species has different patterns of distribution and behavior [37], [42], lineages that are limited to some vector species are restricted only to bird species that this vector could bite. Moreover, as the number of samples from each bird species was unbalanced, the results of the host range may overestimate cases of host specificity. In the present study, we verified that the Plasmodium lineages presented a broader spectrum of host species than did the Haemoproteus lineages. The evidence that indicated the restriction of Haemoproteus lineages at host level was also observed in community-level studies in Africa and Europe, suggesting that this could be an inherent attribute of the Haemoproteus genus [39], [43]–[44].

Phylogenetic Relationships

The Bayesian phylogenetic inferences of the Plasmodium lineages revealed five major clades without a noticeable association with landscape or avian host family. However, associations among some subclades and particular host families and habitat types could be observed. First, the lineages that compose the most diverse subclade in Clade 1 almost exclusively infect birds from the Atlantic rainforest communities, despite these same bird species dwelling in other habitat types. This pattern suggests that some components of the Atlantic rainforest, such as suitable vectors, could limit lineages from this subclade to this particular habitat. Second, with the exception of the lineage GARUF01, all lineages from the subclade that includes the widespread P. elongatum were found in Turdus species as a host. The pairwise distances between the P. elongatum lineage GRW06 and all other lineages from this subclade ranged from 0.02% to 2.9%, suggesting that these lineages represent different lineages of P. elongatum. These lineages were established in the community after obtaining suitable conditions in Brazilian Turdus as a host species. Third, the majority of Clade 4 lineages were found in multiple bird species, suggesting that the generalist strategy of host exploration should be an inherited component of Clade 4 members instead of being an independently acquired characteristic for each species.

The traditional taxonomy of Heamoproteus, which is based on morphological characters observable by light microscopy, places the Haemoproteus species that infect Columbiformes in the sub-genus Haemoproteus, whereas all other Haemoproteus species should belong to the sub-genus Parahaemoproteus [2]. Recently, molecular evidence has supported the hypothesis that Haemoproteus (Haemoproteus) and Haemoproteus (Parahaemoproteus) can be considered distinct clades, maintaining the proposal of a sub-genus exclusively composed of species that infect Columbiformes [34], [45]. Our phylogenetic tree of Haemoproteus lineages also revealed the occurrence of two distinct clades corresponding to the Haemoproteus and Parahaemoproteus clades (Fig. 2). However, the lineage COTAL01, which belongs to the Haemoproteus clade, was found to infect two individuals of the Coccyzus melacoryphus (Cuculidae) species, a non-Columbiformes bird species. These findings introduce the possibility of the use of non-Columbiformes hosts by members of the sub-genus Haemoproteus (Haemoproteus), although a clear confirmation should include the morphological characterization of these parasites.

We observed that the pairwise genetic distances within the clades varied significantly, and in some cases, the intra-clade variation exceeded the variation among the clades. This pattern of variation hinders the establishment of divergence thresholds between lineage groups and prevents the proposition of putative Plasmodium and Haemoproteus species based on genetic distance values using the cyt-b marker.

Associations among Habitat, Composition and Richness of Parasite Communities

The composition analyses of parasite communities revealed the same pattern presented by the bird sampling: a significant difference between birds of the Atlantic rainforest versus sampled species of birds from the Brazilian savannah and seasonally dry forest communities (Table 2 and Fig. 3). Atlantic rainforests are characterized by a high and constant rainfall and a narrow temperature variation throughout the year due to the influence of the coastal environment. Meanwhile, the Brazilian savannah and seasonally dry forests undergo more variation across the year from the two seasons, wet and dry [46]. Therefore, it is reasonable to hypothesize that the abiotic attributes of Atlantic rainforests may be associated with the differences in the composition of bird communities. Considering that our bird sampling represents a real picture of the composition of the bird communities, the presence of the same pattern for both bird and parasite communities indicates that the composition of a particular bird community affects the composition of malaria parasite lineages, most likely due to events of host specialization that occur over co-evolutionary interactions in a community.

The GLM analyses revealed that areas with a greater diversity of bird species also had high parasite richness, supporting the hypothesis that areas with high bird richness also have high parasite richness by providing more diverse host environments for different lineages of parasites to colonize. However, our sample size per site is positively correlated with lineage richness; thus, the uneven sampling effort among the sites may have influenced the results.

Concluding Remarks

It is noteworthy that most of the lineages were reported here for the first time, indicating the singularity of the parasite community of the sampled areas and revealing the need for studies that address the relationship between these parasites and their hosts. Indeed, only 14 of the 110 lineages identified in this study had been described previously. The identification of previously described lineages is relevant for a better understanding of their spatial distribution and host range; for example, the P. elongatum GRW06 lineage that had already been found in birds from all other continents is described for the first time in South America, infecting three bird species. Moreover, further studies using greater and uniform sampling areas with well-founded evidence of differences in bird richness would be valuable to investigate the effects of bird community structure on the diversity of avian malaria communities and corroborate the results from this study.

Sampling Design

A total of 1,545 DNA samples from birds collected at 10 sites across the southeast Brazilian region were selected for this study (Table 1, Fig. 5). These DNA samples are part of the two bird DNA banks: LBEM DNA bank (maintained by FRS) and Malaria Lab Bird DNA bank (maintained by EMB). Samples from LBEM DNA bank were collected between 2000 and 2006 including samples from ARA, BOC, BRA, CAR, FEL, JEQ, NOV, SAL and SOO sites. Samples from Malaria Lab Bird DNA bank were obtained during 2010, including samples from MAN site. These sites were selected due to their representative bird communities with a variable richness of species comprising three different types of tropical ecosystems: Atlantic rainforest, Brazilian savannah and tropical dry forest [22]. The blood obtained from all birds, which were caught using mist nets, was stored in absolute alcohol in an ultra-freezer at −70C°. DNA extraction was performed with a phenol–chloroform–isoamilic alcohol protocol described in Sambrook et al. [47]. Samples from the LBEM DNA bank were collected between 2000 and 2006, whereas samples from the Malaria Lab Bird DNA bank were obtained during 2010.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Map of southeast Brazil depicting the sampling sites and the delimitation of the highest areas of Espinhaço range.

The names of the sampling sites are coded (ARA = Aracruz; BRA = Brasilândia; BOC = Bocaiúva; CAR = Caratinga; FEL = Felixlândia; JEQ = Jequitinhonha; MAN = Manga; NOV = Nova Lima; SAL = Salto da Divisa).

https://doi.org/10.1371/journal.pone.0057770.g005

The study was approved by the Ethics Committee in Animal Experimentation (CETEA), Federal University of Minas Gerais, Brazil (Protocol #254/2011).

Molecular Analyses

DNA samples were initially screened for the presence of Plasmodium and/or Haemoproteus infections by PCR using primers 343F (5′GCTCACGCATCGCTTCT3′) and 496R (5′GACCGG-TCATTTTCTTTG3′) according to the protocol described by Fallon et al. [48]. The presence of a 195-bp band in acrylamide gels indicates the presence of Plasmodium and/or Haemoproteus infection. Subsequent to parasite screening, a 524 bp fragment of the mtDNA cytochrome b gene from the infected individuals was amplified by a nested-PCR using primers HaemNFI (CATATATTAAGAGAAITATGGAG) and HaemNR3 (ATAGAAAGATAAGAAATACCATTC) in a first amplification; and HaemF (ATGGTGCTTTCGATATATGCATG) and HaemR2 (GCATTATCTGGATGTGATAATGGT) in a second amplification, following protocols described by Hellgren et al. [49]. The PCR products were purified using a solution of 20% polyethylene-glycol 8000 and 2.5 M NaCl according to the methods of Sambrook et al. [47]. After purification, the PCR products were sequenced in both directions using the BigDye Terminator Kit v3 (Applied Biosystems, Foster City, CA, USA) using an ABI3100® automated sequencer (Applied Biosystems, Foster City, CA, USA).

The sequences were assembled and checked for quality using Phred v.0.20425 [50]–[51] and Phrap v.0.990319 [52]. The assembled chromatograms were carefully checked and edited using Consed 23.0 [53]. Sequences showing double peaks were examined for multiple infections by cloning (TOPO-cloning kit, Invitrogen®) and sequencing. We sequenced six clones from each sample for which we observed a likely co-infection. Sequences were aligned using the Clustal W algorithm implemented in MEGA version 5 [54]. As there is no consensus on how to delimit haemosporidian species based on cyt-b sequences [55], we considered each haplotype as an independent lineage and our unit of richness for further analyses. The genus of each lineage was inferred by the closest sequence matches in Genbank using an NCBI nucleotide Blast search. In an attempt to assign the sequences to describe parasite lineages, we compared the sequences with records in the Genbank and MalAvi databases [56], which contain cyt-b data for most of published avian haemosporidian parasite lineages. Observed lineages that were not present in the MalAvi database were considered new lineages. All sequences were deposited in Genbank (see Table S1).

Phylogenetic and Statistical Analyses

Bayesian analyses were implemented to infer the phylogenetic relationships among cyt-b lineages. MrBayes version 3.1.2 [57] was used to run two Markov chains simultaneously for 3 million generations that were sampled every 100 generations. The first 7500 trees (25%) were discarded as a “burn-in” step and the remaining trees were used to calculate the posterior probabilities. The Plasmodium and Haemoproteus lineages were analyzed separately because there is no consensus about the monophyly of both genera [34], [58]. A sequence of P. malariae was used as an out-group based on recent evidence suggesting that the Plasmodium genus is polyphyletic, and Plasmodium species that infect mammals form a sister clade with the clade that harbors the groups studied here [58]. Sequence divergence within and among the major Plasmodium and Haemoproteus clades was calculated using uncorrected-P distance in MEGA version 5 [54].

For statistical analyses, we first used a non-metric multidimensional scaling (NMDS) model to search for overall differences in parasite lineage compositions between the following habitats: Atlantic rainforest, Brazilian savannah and tropical dry forest. We also used an analysis of similarities (ANOSIM) to test the significance of the differences found by the NMDS. The composition analyses were performed using the software PAST [59]. We performed similar composition analyses for the bird samples to identify the possible effects of bird sampling in the NMDS results for parasite lineages.

Next, we applied generalized linear models (GLMs) to identify possible effects of the type of habitat and bird richness (explanatory variables) on the prevalence of infection and lineage richness (response variables). All GLMs were submitted to residual analyses to evaluate the adequacy of the error distribution. Minimum adequate models were generated by a stepwise omission of non-significant terms. The GLMs were performed with the software R [60]. Additionally, for prevalence analyses, we excluded data from bird species with less than 5 observations to avoid sampling bias.