INTRODUCTION
The discovery of a malaria parasite in human blood by Laveran in 1880 was followed by recognition that a substantial diversity of related parasite species is found in reptile, bird, and mammal hosts almost worldwide (reviewed by Garnham, 1966). Debate quickly emerged on how to place these species into genera, as well as over the relationships of species within each genus. The central issue for over a century has been the choice of appropriate characters to define taxa of the parasites from species to families (Corradetti et al. 1963; Garnham, 1966; Valkiunas, 2005). Modern systematic biology requires that higher taxa represent monophyletic groups of species, and the characters used to define such taxa must be phylogenetically informative. That is, taxa must be defined based on shared derived character states (synapomorphies). The characters traditionally used to define subgenera of Plasmodium, for example, include morphology seen under the light microscope, details of life-cycles, insect and vertebrate host range and, more recently, ultrastructural morphology (Garnham, 1966; Laird, 1998; Valkiunas, 2005). However, the phylogenetic significance of these characters has long been problematic (Corradetti et al. 1963; Garnham, 1966; Manwell, 1936) and recent molecular studies have recalled the issue (Escalante et al. 1998; Perkins and Schall, 2002).
Based on traditional characters, fully 11 subgenera are now considered valid for Plasmodium, with a new subgenus being erected only within the past decade (Garnham, 1966; Valkiunas, 1997). A recent analysis of gene sequences (the cytochrome b gene from the mitochondrial genome) supported the monophyly of each of the 3 named subgenera of Plasmodium found in mammal hosts (Laverania of humans/chimpanzees, Plasmodium in other apes, humans, and monkeys, and Vinckeia in rodents) (Perkins and Schall, 2002; Perkins et al. 2006). However, the status of the primary diversity of Plasmodium, which includes the species exploiting reptiles and birds as the vertebrate hosts, was unresolved. Kissinger et al. (2002) used sequences of the small-subunit ribosomal RNA gene to examine relationships of 5 Plasmodium species from avian hosts and concluded that the currently accepted subgenera of avian malaria parasites are not phylogenetically valid (the subgenera do not represent independent evolutionary groups of species). Molecular phylogenies also cast doubt on the value of traditional characters used to define higher taxa of malaria parasites (Escalante et al. 1998; Perkins and Schall, 2002). That is, characters such as blood cell type exploited, shape and size of schizonts and gametocytes in the blood, periodicity of infection in vertebrate host, and number of merozoites in a mature schizont may often be homoplasious (character states evolved independently in different lineages). In contrast, a detailed study of several species concepts (morphology, genetic, and phylogenetic) for an assemblage of avian haemosporidian parasites revealed that morphology alone usually assigns parasite lineages to monophyletic groups that may represent species (Martinsen et al. 2006).
We ask here if the recognized subgenera of avian Plasmodium, erected primarily on morphology, represent monophyletic groupings of species. We use sequences for 2 genes from the parasite's mitochondrial genome, cytochrome b and cytochrome oxidase I, to recover a phylogeny of Plasmodium identified to subgenera based on morphology seen in blood smears. The infections were identified from a large sample of birds collected primarily from Vermont and California, USA. This is the first 2-gene phylogenetic study of avian Plasmodium. Our goal is not to propose a taxonomic revision of the genus Plasmodium, but to determine if the often-striking morphological characteristics used to define the subgenera are phylogenetically significant. If these characters do not successfully define monophyletic groups, this would suggest they are under differing selective pressure within parasite evolutionary groups, an intriguing and significant finding. The results thus cast light on the evolutionary significance of morphological characters used for more than a century to define groups of Plasmodium species.
MATERIALS AND METHODS
Birds were collected from sites throughout Vermont (USA) and a site in northern California (USA) near the town of Hopland in Mendocino county (University of California Hopland Research and Extension Center) using mist-nets and potter traps. After each bird was identified, a blood sample was obtained by puncture of the brachial vein and used to produce a thin smear to be fixed in absolute methanol and blood dots dried on filter paper to be stored frozen with silica gel beads. All sampling was conducted under governmental permits and University of Vermont animal care guidelines. Several other samples were provided by colleagues, 1 from Uganda that provided sufficient material for proper diagnosis, and 2 samples from Israel with highly confident identification (Martinsen et al. 2006). Sequence data for the 2 genes from P. relictum of Hawaiian birds were extracted from GenBank (AY733090). This is a very well-characterized sample of P. relictum and thus identification to species is reliably corroborated (Beadell and Fleischer, 2005). The subgenus Bennettinia presently is known to include only P. juxtanucleare of domestic chickens (Valkiunas, 1997), and is represented here by 3 samples collected in Vietnam. These samples are confidently assigned to P. juxtanucleare because they are well characterized by researchers at the Pasteur Institute. Data from GenBank for another isolate of P. juxtanucleare from Japan are also included in the analysis (YP717012).
Smears were treated with Giemsa stain (1[ratio ]10, pH 7·0, 50 min) and the entirety of each slide examined at 1000× magnification. Only infections containing substantial numbers of schizonts and gametocytes were used to determine subgenus. Most Plasmodium infections revealed parasitaemia too low for confident diagnosis to subgenus, and were discarded from the study. Morphological characters used to identify the subgenera are presented in Table 1. Only characters apparent on the stained blood smears were available for the study. Although some of these characters overlap between subgenera, a combination allowed confident identification.
Table 1. Currently recognized subgenera of Plasmodium from avian hosts with key defining characters seen in stained blood smears (Example species for each subgenus are also provided. Diagnostic characters were extracted from the type description for each subgenus, Valkiunas (2005), Garnham (1966), and Laird (1998).)
DNA was extracted from dried blood dots with the DNeasy kit (Qiagen, USA). A final 614 bp segment of the cytochrome b gene (cyt b) and a 982 bp segment of the cytochrome oxidase I gene (coI) were amplified using nested PCR. Outer reactions for cyt b gene used primers DW2 (5′-TAA TGC CTA GAC GTA TTC CTG ATT ATC CAG-3′) and DW4 (5′-TGT TTG CTT GGG AGC TGT AAT CAT AAT GTG-3′) (Escalante et al. 1998; Perkins and Schall, 2002). Outer reactions for the coI gene used primers CoIoutF (5′-CTA TTT ATG GTT TTC ATT TT ATT TGG TA-3′) and CoIoutR (5′-AGG AAT ACT AGG CAT TAC ATT AAA TCC-3′) (Perkins et al. 2006). Both outer reactions contained 2 μl of extraction product, 1 μl of each 10 μM primer, and a Ready-to-Go PCR bead (Amersham, USA). An initial denaturation for 4 min at 94 °C was followed by 35 cycles of 94 °C for 30 sec, 60 °C for 30 sec, and 68 °C for 50 sec. The inner reactions contained 1 μl of the outer product and used primers DW1 (5′-CAT ATC CTA AAG GAT TAG AGC TAC CTT GTA A-3′) and DW3 (5′-TGC TGT ATC ATA CCC TAA AG-3′) (Perkins and Schall, 2002) for cyt b, and CoIinF (5′-ATG ATA TTT ACA RTT CAYGGW ATT ATT ATG-3″) and CoIinR (5′-GTA TTT TCT CGT AAT GTT TTA CCA AAG AA-3′) (Perkins et al. 2006) for coI. PCR conditions for inner reactions were an initial step of 94 °C for 1 min followed by 40 cycles of 94 °C for 20 sec, 52 °C for 20 sec, and 68 °C for 30 sec, and a final step of 7 min at 68 °C. A negative control was used for each PCR run, and no contamination was detected by gel electrophoresis of the PCR products.
After purification (ExoSAP-It, USB, USA), the PCR products were sequenced using BigDye Terminator v3.0 and analysed on an ABI Genetic Analyzer (ABI, USA) at the University of Vermont Cancer Center. The cytochrome b gene fragment was sequenced using the DW1 primer, and the cytochrome oxidase I gene fragment with the CoIinF primer. Sequences were edited using Sequencher (Genecodes, USA), aligned by eye in PAUP* 4.0b10 (Swofford, 2002), and examined for the presence of stop codons using MacClade version 4.03 (Maddison and Maddison, 2001). Indels were not observed for either gene, so alignment was unambiguous. If inspection of any electropherogram revealed ambiguous base calls, the sample was re-amplified, sequenced, and edited. Continued presence of ambiguous bases suggested a mixed-species infection, and these samples were discarded from the study.
Phylogenetic analysis used 2 rodent malaria parasites, Plasmodium berghei and P. chabaudi as outgroup taxa because a previous study (Perkins and Schall, 2002) and our confirming analysis demonstrated that these species are closely related to but not contained within the avian Plasmodium clade. Sequence data for the rodent parasites were obtained using the same protocols described above for samples provided by R. Carter (University of Edinburgh). Phylogenetic reconstruction was conducted under both parsimony and likelihood frameworks and hypothesis testing was carried out using the Shimodaira and Hasegawa test (Shimodaira and Hasegawa, 1999).
Both unweighted and weighted parsimony analyses were performed. Unweighted maximum parsimony analysis involved 10000 random stepwise-addition heuristic replicates as implemented in PAUP* 4.0b10 (Swofford, 2002). A site-specific rate weighting scheme was also employed, a method shown to be useful when gene conservatism, unequal nucleotide frequencies, and differences in substitution rates are encountered among and within datasets, such as for the cyt b and coI genes used in our phylogenetic analysis (data not shown) (Kjer et al. 2001). Each character from the 2-gene dataset was assigned to one of 5 rate classes based upon a fast heuristic bootstrap analysis (1000 replicates) in PAUP* (Kjer et al. 2001). A heuristic search with 10000 random stepwise-addition replicates was then performed in PAUP* with each character weighted inversely to its rate of change. Nodal support values were generated using 10000 heuristic bootstrap replications under both the unweighted and site-specific rate weighting schemes.
Phylogenetic analysis using likelihood criteria was conducted using the Markov Chain Monte Carlo (MCMC) sampling regime in MrBayes (MrBayes v3.0b4; Hulsenbeck and Ronquist, 2001). The most appropriate model of evolution was selected separately for each gene using the Akaike Information Criterion (ModelTest version 3.06; Posada and Crandall, 1998). A general time-reversible model with a gamma distribution for variable sites and proportion of sites as invariable (GTR+Γ+I) was chosen for both cyt b and coI. Starting parameters used for the Bayesian analysis included a starting tree estimated from neighbour-joining methods in PAUP*. Additionally, the data were partitioned into the 2 genes, with a separate model of evolution (GTR+Γ+I) designated for each data partition. The MCMC sampling regime was run twice independently for 1 million generations with sampling every 100 generations (MrBayes v3.0b4). A total of 10001 trees resulted from each analysis, with the first 1000 trees discarded as burn-in or suboptimal trees. The 9001 remaining trees from each analysis were then combined (N=18002) and used to calculate posterior probability values.
To compare the topology resulting from our analyses to the hypothesis that the subgenera of avian Plasmodium are monophyletic, hypothesis testing was carried out using maximum likelihood analysis and the Shimodaira and Hasegawa test in PAUP*. A GTR+Γ+I model of evolution was appropriated to the combined gene dataset using the Akaike Information Criterion of ModelTest v3.06 (Posada and Crandall, 1998) and a likelihood heuristic search involving 100 random stepwise-addition replicates performed. The resulting maximum likelihood tree was then used to construct an alternative topology by branch rearrangement in MacClade, such that each subgenus was constrained to a clade (and thus to monophyly). This alternative topology was then loaded as a constraint for a maximum likelihood analysis identical to that above. In this additional likelihood analysis, only trees in which the subgenera were monophyletic were evaluated. The best tree obtained from the constrained maximum likelihood analysis was then compared to the observed maximum likelihood tree using the Shimodaira and Hasegawa test including 1000 bootstrap replicates with RELL approximation in PAUP* (Shimodaira and Hasegawa, 1999).
RESULTS
A total of 2200 birds from Vermont and 432 birds from California were screened by light microscopy for malaria parasites, with 204 (Vermont) and 6 (California) found infected with Plasmodium. Most of these infections (N=180) were low-grade parasitaemia and/or lacked one of the developmental stages, and so were discarded from the study. In addition, 5 infections presented ambiguous sequences due to mixed infections of the genera Plasmodium and Haemoproteus, and these samples were discarded (because genus-specific primers have not yet been developed for the coI gene). This left 25 infections from Vermont and California from a range of passerine families that could be confidently placed into subgenera. Addition of 2 samples from Israel, 1 from Uganda, 3 from Vietnam and the GenBank sequences available for the Hawaiian P. relictum and a Japanese isolate of P. juxtanucleare resulted in a total of 33 avian Plasmodium samples.
Three samples were identified to the subgenus Huffia, 9 samples to the subgenus Haemamoeba, 3 samples to the subgenus Giovannolaia, 4 samples to the subgenus Bennettinia, and 14 samples to the subgenus Novyella. For each of these samples and the 2 outgroup rodent Plasmodium taxa, a total of 1596 nucleotides from the 2 genes were included in the analysis; sequences generated in this study are deposited in GenBank (Accession numbers EF011166–EF011231).
The unweighted maximum parsimony heuristic search resulted in 6 equally parsimonious trees that differed only in the placement of 2 Novyella samples within the major Novyella clade and the order of appearance of identical haplotypes within clades. The weighted parsimony analysis resulted in 3 equally parsimonious trees and a consensus tree identical to that estimated by the unweighted analysis. The strict consensus phylogram summarizing the results from the weighted parsimony analysis including bootstrap support values, is presented in Fig. 1. Bayesian MCMC analysis estimated an almost identical topology, with only slight rearrangement of 2 Novyella samples within the major Novyella clade. Posterior probability values from the Bayesian MCMC analysis are provided in Fig. 1. The likelihood analysis resulted in a topology that differed only in the placement of 1 Novyella sample within the major Novyella clade.
Fig. 1. Phylogenetic relationships of the avian subgenera of Plasmodium based upon 614 nucleotides of the cytochrome b gene and 982 nucleotides of the cytochrome oxidase I gene. A weighted maximum parsimony consensus phylogram (N=3 trees) is presented, with bootstrap values followed by posterior probabilities for each node. Posterior probability values are not provided for nodes recovered only by maximum parsimony analysis as indicated by asterisks. Locations for samples collected outside of Vermont, USA, are provided. Outgroup taxa include the rodent malaria parasites, P. berghei and P. chabaudi.
The 3 samples identified to the subgenus Huffia had identical sequence data for both genes. These were found in a House Finch (Carpodacus mexicanus) and 2 Song Sparrows (Melospiza melodia), all sampled in Vermont. These samples fell onto a well-supported long branch, and thus likely represent a distinct clade. All samples identified to the subgenus Haemamoeba fell into a well-supported clade, with bootstrap support and posterior probability values of 96 and 99% respectively. This Haemamoeba clade includes 2 samples identified as P. relictum, 1 from Israel and 1 from Hawaii. Most samples identified to the subgenus Novyella fell into a well-supported clade according to Bayesian analysis, including samples from Vermont and California. Two other samples, with sequences differing by only a single base pair, were identified as the subgenus Novyella from 2 Red-winged Blackbirds (Agelaius phoeniceus) and fell onto a long well-supported branch basal to all other avian Plasmodium in the study. The 3 samples of Bennettinia from Vietnam produced identical sequences and differed by only a single base pair from the Japanese sample extracted from GenBank. These samples fell on a long branch basal to most other Plasmodium included in the study. Thus, the morphologically distinctive subgenus Bennettinia appears to form a valid clade. Although only 3 samples were identified to the subgenus Giovannolaia, these fell within 3 distant clades on the phylogeny, and clearly did not form a monophyletic group. One from Vermont (from a Northern Waterthrush, the warbler Seiurus noveboracensis) fell as a sister taxon to the Haemamoeba clade, and the other 2 fell within a large clade that contained most of the ‘Novyella’ samples (from an American Robin, Turdus migratorius, from Vermont, and from the Ugandan African Fish Eagle, Haliaeetus vocifer).
The subgenera Huffia, Haemamoeba, and Bennettinia are monophyletic by all 3 methods of phylogenetic inference. As representatives of the subgenera Novyella and Giovannolaia were interspersed only with each other and not contained within nodes defining the subgenera Huffia, Haemamoeba, and/or Bennettinia, a single constraint tree was constructed in which both of these subgenera were monophyletic. The hypothesis of monophyly for Novyella and Giovannolaia was then tested against the observed topology using the Shimodaira and Hasegawa test. Results of this test indicate that the constraint topology was significantly different than the observed topology by likelihood criteria (P<0·0001). Thus, the constraint topology displaying monophyly for the subgenera Novyella and Giovannolaia was seen as a significantly worse topology according to likelihood analysis of our 2-gene mitochondrial dataset.
DISCUSSION
Early efforts to understand the diversity of malaria parasites centred on defining the evolutionary relationships of the parasites using morphological characters from stages found in both the vertebrate and insect host. Also considered important were life-history features (variation in details of the life-cycle) and the taxon of both vertebrate and insect host (reviewed by Garnham, 1966; Valkiunas, 2005). The phylogenetic significance of these characters was questioned long ago by Manwell (1936) when he proposed that only genetic data would reveal the true relationships of the parasites. When Corradetti et al. (1963) reviewed the avian Plasmodium and erected 2 new subgenera (Novyella and Giovannolaia) they viewed the use of morphological characters as an imperfect guide to the true evolutionary relationships of Plasmodium species. The early doubts of Manwell, Garnham, Corradetti and others are now mirrored by findings based on new phylogenies recovered using gene sequence data. These results suggest that life-history traits of the biology of malaria parasites have evolved independently in phylogenetically distinct lineages (Escalante et al. 1998; Perkins and Schall, 2002; Perez-Tris et al. 2005, but see Valkiunas et al. 2005). For example, P. giganteum, a parasite of west African lizards, that produces enormous schizonts in erythrocytes containing ~100 merozoites, is not related to other Plasmodium giants, but is the closest sister to the relatively tiny P. agamae (Perkins and Schall, 2002).
After a large number of Plasmodium species were described (now ~180) from mammals (primarily from primates and rodents), birds (from a very broad range of families), and lizards (also from many families), parasitologists erected an array of subgenera in an attempt to maintain the genus Plasmodium as a single taxon (Corradetti et al. 1963; Garnham, 1966; Valkiunas, 2005). Previous molecular phylogenies supported the monophyly of the subgenera of Plasmodium of mammals, but the relationships of parasites of birds and reptiles remained unresolved (Perkins and Schall, 2002; Beadell et al. 2004; Szymanski and Lovette, 2005). Kissinger et al. (2002) presented a phylogeny based on an rRNA gene for several isolates of avian malaria identified to subgenus, and questioned the value of the subgenera erected for the bird Plasmodium. Our results are less pessimistic than those of Kissinger et al. (2002) because the new analysis supports 3 of the 5 named subgenera of avian Plasmodium (below). We approached the problem by including a larger sample of infections, using sequence data from 2 genes, and analysing the data with a spectrum of methods, including hypothesis testing.
A major challenge encountered in our study was logistical – finding infections with sufficient parasites of all stages needed to score the morphological characters that define the subgenera. Only 14% of the Plasmodium infections contained a full range of immature and mature schizonts and gametocytes that were required for comparison to literature descriptions of the parasites. Valkiunas (2005) also casts a warning that using only a few parasites or a single life-stage will result in many errors in identification. Therefore, we used only infections with plentiful parasites of all stages, and examined each slide completely at least 3 times. Any other protocol will present questionable results.
The study resulted in a phylogeny that was well resolved and with very similar topology by all methods of analysis. This phylogeny suggests a new hypothesis for the relationships between the parasites and offers some insight into the venerable issue of the value of morphology in systematic studies of Plasmodium. Five clades are apparent. One clade included 3 samples, apparently from the same Plasmodium species, with the Huffia morphology of variable schizont size and shape, plentiful cytoplasm, elongated gametocytes, and immature erythrocytes as host cells. Concluding that Huffia represents a monophyletic group is premature, but the samples fell onto a well-supported long branch as would be expected if Huffia is indeed a distinct clade. A second clade included parasites matching the morphology of Haemamoeba, with round or oval schizonts and gametocytes that often displace the host cell nucleus. Within this clade were samples identified definitively as P. relictum, 1 from Israel and 1 from Hawaii. P. relictum has been implicated in the catastrophic decline of the endemic Hawaiian avifauna (van Riper et al. 1986; Atkinson and van Riper, 1991). The P. relictum from Hawaii falls closely with the sample identified as P. relictum from Israel. The third clade includes most samples identified to Novyella and some Giovannolaia, and contains species with diverse morphology, but always elongated gametocytes. The morphological distinctions between Novyella and Giovannolaia are not clear, and have been viewed as ambiguous (Corradetti et al. 1963; Seed and Manwell, 1977; Laird, 1998). Indeed, upon erecting the subgenera Novyella and Giovannolaia, Corradetti et al. (1963) admitted a difficulty in assignment of some parasite species to one of these 2 subgenera. Our results clearly demonstrate that these subgenera do not form 2 monophyletic clades, and the morphological characters differentiating the 2 subgenera do not appear to be phylogenetically informative. The fourth clade includes all 4 samples identified to the monospecific subgenus Bennettinia, and this clade fell basal to almost all other Plasmodium in the study. Bennettinia was erected by Valkiunas (1997) for P. juxtanucleare, a distinctive small parasite that usually adheres to the host cell nucleus. The last clade was identified from only 2 samples with the Novyella morphology, but falling basal to all avian Plasmodium. An enigmatic single sample, with the morphology of Giovannolaia, fell as a sister taxon to the entire subgenus Haemamoeba. The currently recognized subgenus Giovannolaia appears invalid.
If this phylogenetic hypothesis is correct, it suggests that the characters used for over a century by parasitologists to define relationships among Plasmodium taxa are not always informative for systematic studies. However, the adaptive significance of the variation in morphology and life-history traits (such as use of the type of host cell) remains an intriguing open question. For example, the striking morphology of Haemamoeba species, including large round gametocytes and prominent host cell nucleus displacement, represents an important variant in how the parasite grows and functions within its host cell, and seems to have evolved once. Similarly, ready occupation of immature red blood cells indicates the unique evolutionary lineage of the Huffia samples identified in this study. P. juxtanucleare has a morphology similar to the Haemamoeba species (round schizonts and gametocytes), but the parasite is very small, and often sticks to the host cell nucleus. Valkiunas (1997) noted the unusual morphology of this species and its very wide geographical distribution in the domestic chicken and its relatives, and erected the subgenus Bennettinia to include only P. juxtanucleare. His analysis is supported here. The unusual morphology of P. justanucleare appears to represent a unique evolutionary event in the evolution of avian malaria parasites. In contrast, the subtle variation in morphology seen as important in distinguishing parasites among the subgenera Novyella and Giovannolaia appears to be plastic over evolutionary time and does not concord with distinct clades. The character of elongate gametocytes may be biologically important, although its significance to the identification of evolutionarily independent groups needs to be further explored because the subgenera Huffia, Novyella, and Giovannolaia all display this trait and the subgenera Novyella and Giovannolaia do not represent monophyletic groups as indicated by our analyses. Molecular phylogenies, therefore, can reveal more than simply the relationships of the diversity of Plasmodium species, but can also focus attention on the question of adaptive significance for the morphological and life-history differences among species.
We thank colleagues who generously provided samples for this study: R. Eisen (California samples), I. Paperna (Israel samples), J. Stuht (Uganda sample), R. Paul (P. juxtanucleare), and R. Carter (rodent Plasmodium samples). Discussions with C. W. Kilpatrick, P. O'Grady, S. Perkins, and K. Kjer improved the analysis. We greatly appreciate the training and assistance of W. Barnard with bird sampling methods. The research was funded by grants from the Morris Animal Foundation to J.J.S., Vermont Genetics Network through the NIH BRIN program and the Theodore Roosevelt Memorial grant from the American Museum of Natural History to E.S.M., and the University of Vermont Helix and URECA! programs to J.L.W.
