The primate malaria parasites Plasmodium malariae, Plasmodium brasilianum and Plasmodium ovale spp.: genomic insights into distribution, dispersal and host transitions

During the twentieth century, there was an explosion in understanding of the malaria parasites infecting humans and wild primates. This was built on three main data sources: from detailed descriptive morphology, from observational histories of induced infections in captive primates, syphilis patients, prison inmates and volunteers, and from clinical and epidemiological studies in the field. All three were wholly dependent on parasitological information from blood-film microscopy, and The Primate Malarias” by Coatney and colleagues (1971) provides an overview of this knowledge available at that time. Here, 50 years on, a perspective from the third decade of the twenty-first century is presented on two pairs of primate malaria parasite species. Included is a near-exhaustive summary of the recent and current geographical distribution for each of these four species, and of the underlying molecular and genomic evidence for each. The important role of host transitions in the radiation of Plasmodium spp. is discussed, as are any implications for the desired elimination of all malaria species in human populations. Two important questions are posed, requiring further work on these often ignored taxa. Is Plasmodium brasilianum, circulating among wild simian hosts in the Americas, a distinct species from Plasmodium malariae? Can new insights into the genomic differences between Plasmodium ovale curtisi and Plasmodium ovale wallikeri be linked to any important differences in parasite morphology, cell biology or clinical and epidemiological features?


Background
In The Primate Malarias (1971), by Coatney et al. [1], detailed species comparisons are presented based on descriptive morphology of both blood and mosquito stages, the geographic distribution of each parasite and certain features readily measurable in induced human infections, including the estimated duration of the liver-stage, time to symptoms and fever periodicity. Much of this work was performed in prison inmates in Georgia, USA. In this paper, fifty years since, the focus on the geographic, genomic and genetic characteristics of four primate malaria species-one currently regarded as zoonotic in South American monkeys, Plasmodium brasilianum, and three malaria parasites of Homo sapiens, namely Plasmodium malariae, Plasmodium ovale curtisi and Plasmodium ovale wallikeri. An exhaustive bibliography of reported identification of these species since 1890, across the globe and in different primate hosts, will also be presented.
Over the last two decades, the analytical techniques of evolutionary biology and the task of reconstructing phylogenetic relationships within the genus have benefited greatly from the explosion in genomic data available for malaria parasites, and the now well-established practise of non-invasive faecal sampling of parasite genomic material from the faeces of wild primates [2]. This wealth of data provides new understanding of diversity both within and among the primate-infecting Plasmodium species, and points to the importance of transitions into new primate hosts. These transitions are gateways to the radiation of parasite species, but also act as genetic bottlenecks, as evidenced by reduced diversity among parasites in the new host [2,3].
Among the homophilic species considered of clinical importance, a range of life history and transmission strategies are evident, and each of these strategies have their equivalent counterparts among the parasites of living simian hosts, and those of Pan and Gorilla. Thus, the majority of evolution leading to these diverse life histories occurred in the parasite lineages of non-human primates in the evolutionary past. However, as with Plasmodium knowlesi, the zoonotic potential of P. brasilianum shows that host transition can be a dynamic process operating over an extended time period, rather than a singular event, and understanding this in the present is essential to maintain effective malaria elimination strategies world-wide.

History & discovery
The first report of P. brasilianum is based on a finding in the blood of a bald uakari (Cacajao calvus) imported from the Brazil Amazonas region to Hamburg, Germany in 1908 [4]. Initial studies reported that P. brasilianum closely resembles P. malariae, and to be a relatively common parasite of New World monkeys in Panama and Brazil (reviewed in [1]).

Distribution and known non-human primate hosts
Historically, natural infections of P. brasilianum were reported in various primates in Central and Southern America-Panama, Colombia, Venezuela, Peru, and Brazil. The spectrum of primate hosts (incl. sequence confirmed reports) is given in Table 1 [5][6][7][8][9][10][11][12], indicating that P. brasilianum has promiscuous host-specificity compared to other malaria parasites. Moreover, natural infections in humans have been reported from Venezuela [13].

Genomic studies of Plasmodium brasilianum
Plasmodium brasilianum is a parasite thought to be closely related to P. malariae, and blood-stage infections of the two species present a morphologically identical picture, with discrimination determined by the host, monkey or human, respectively. The few molecular epidemiological studies reported so far have shown that P. brasilianum and P. malariae infections are almost indistinguishable genetically. Sequencing studies of the gene coding for the circumsporozoite protein (csp) appear not to differentiate the identity of the two parasites [14][15][16]. Similar, studies involving the merozoite surface protein-1 (msp1), the ssrRNA small subunit (18S) of ribosomes and the mitochondrial gene cytochrome b (cytb), have identified sequences that were 100% identical or that had only a few randomly distributed single nucleotide position differences [7,13,[15][16][17][18]. Further, the close genetic resemblance of these parasites has been observed across studies in Brazil, Venezuela, Costa Rica, Peru, Colombia and French Guiana from infected humans, monkeys and mosquitoes [7-9, 11, 12, 15-18]. Under conditions of close contact, as shown in Yanomami people and monkeys species in the Venezuelan Amazon, both humans and non-human primates shared quartan parasites without any host specificity that are genetically identical in target candidate genes [13].
A small study using microsatellite genotyping showed that in 14 P. malariae isolates from infected individuals from the Brazilian Atlantic forest, all isolates had identical haplotypes, while in one mosquito sample from the same region a different haplotype was found [19]. In the same study, three P. brasilianum isolates from non-human primates sampled from a different region (Amazonia) were analysed, and diverse haplotypes were observed. Unfortunately, across all such studies to date only a small number of samples have been compared at only a few genetic loci. To understand the degree of similarity among P. brasilianum and P. malariae parasites, a comprehensive analysis of whole genome sequencing data is necessary, using many more parasites obtained from different hosts, across a range of geographic regions. Only one draft reference genome of P. brasilianum is available [20]. Similarly, only a few genomes are available for P. malariae, sourced from Africa and Asia, and none from South America [8,[20][21][22]. The apicoplast and mitochondrion genomes of P. brasilianum are indistinguishable from those of the P. malariae reference genome [20,23], but further comparative analysis of nuclear genomes is needed to fully understand the status of these two species. This is made difficult by the scarcity of whole genome data, so it remains an open question whether these parasites are variants of a single species that is naturally adapted to both human and New World monkey hosts, and freely circulates between them. Related to this, it is also difficult to infer the direction of the cross-species transfer. Nevertheless, the similarity of these parasites suggests that monkeys can act as Bolivia, Brazil, Peru JX045642 [12] reservoirs of P. malariae / P. brasilianum, and this must be considered in control and eradication programmes.

Plasmodium malariae
History & discovery; epidemiology and disease As Collins and Jeffery relate [24], P. malariae was named by Grassi and Feletti in 1890, following the observations of Golgi in 1886, who noted the existence of malaria parasites with either 48 h or 72 h cycles of fever, the latter subsequently being recognized as characteristic of P. malariae infections. This slow-growing species is widely distributed across the tropics and sub-tropics, with often asymptomatic infections characterized by low parasitaemia and a recognized ability to persist in a single host for years or decades [25,26]. There is evidence that P. malariae can survive combination therapies used for treating acute P. falciparum malaria, and may present as a post-treatment recrudescence in P. falciparum patients [27][28][29]. Clinical malaria caused by P. malariae rarely progresses to severe, complicated or life-threatening illness, although the literature contains consistent reports of mortality due specifically to either glomerulonephritis or severe anaemia in small children with chronic infections [30].

Distribution and abundance
Plasmodium malariae is a cosmopolitan parasite distributed in sub-Saharan Africa, South-East Asia, western Pacific islands, and Central and South America [24]. Formerly this parasite was also present in the southern parts of the USA, Argentina, Bhutan, Brunei, South Korea, Morocco, Turkey, and parts of Europe where malaria was eradicated [31][32][33]. The distribution of this parasite is variable and patchy, and limited to particular mosquito vectors (sporogony needs a minimal temperature of 15 °C), yet autochthonous P. malariae cases have been documented from much of the tropics and sub-tropics ( Fig. 1; Table 2) . Assessment of the abundance of P. malariae is difficult because this parasite has been neglected by researchers, and studies differ (e.g. symptomatic patients vs. population studies; Table 2). Some epidemiological studies reported a high prevalence (15-30%) in Africa, Papua New Guinea, and the Western Pacific, in contrast to scanty observations (1-2%) from Asia, the Middle East, Central and Southern America [144]. However, with the advent of molecular diagnostic techniques this parasite species has been reported more frequently, being found in regions where it was not previously thought be present (e.g. Bangladesh), more commonly observed in mixed infections with P. falciparum [24], and identified as recrudescent infections in historical cases from areas such as Greece, formerly endemic for malariae malaria, but since having eliminated contemporary transmission of the disease [145].

Genomic studies of Plasmodium malariae
Large-scale genomic studies of the neglected malaria parasites and zoonotic species have been difficult to date, limited by infections having low parasite densities and being mixed with other Plasmodium species, thereby making it difficult to obtain sufficient parasite DNA to perform whole genome sequencing. For P. malariae, the first partial genome using next-generation sequencing was produced from CDC Uganda I strain DNA [22,146]. A subsequent study generated a more complete reference using long-read sequencing technology from DNA of the P. malariae isolate PmUG01, from an Australian traveller infected in Uganda [22,23]. Additional genomic data from short-read Illumina data of travellers' isolates from Mali, Indonesia and Guinea, and one patient in Sabah, Malaysia, were also reported by Rutledge et al. Analysis of these genomes revealed that around 40% of the 33.6 Mbp genome (24% GC content), particularly in subtelomeric chromosome regions, is taken up by multigene families, as seen in P. ovale species [22,25]. The P. malariae genome displays some unique characteristics, such as the presence of two large families, the fam-l and fam-m genes, with almost 700 members [22,23]. Most of these genes encode proteins with a PEXEL export signal peptide and many encode proteins with structural homology to Rh5 of P. falciparum, the only known protein that is essential for P. falciparum red blood cell invasion [147]. These observations suggest that the fam-l and fam-m gene products may also have an important role in binding to host ligands. Other gene families, such as the Plasmodium interspersed repeat (pir) loci that are present in many species in the genus, including in Plasmodium vivax (~ 1500 vir genes), are present in the P. malariae genome. Of the 250 mir genes identified, half are possible pseudogenes. Products of the pir genes are predicted to be exported to the infected erythrocyte surface and may have a role in cell adhesion. Like pir genes, SURFIN proteins are also encoded in the P. malariae genome at around 125 loci, much greater than the number present in P. falciparum (ten) or P. vivax (two). Another unique feature of the P. malariae genome is the presence of 20 copies, in a single tandem array, of the P27/25 gene, a sexual-stage cytoplasmic protein with a possible role in maintaining cell integrity. P27/25 is encoded by a single copy gene in all other species evaluated to date [23,25]. The sequences of an additional eighteen P. malariae genomes from Africa and Asia have recently been reported [21]. These were derived directly from patient isolates, using a selective whole genome DNA amplification (SWGA) approach to increase the relative abundance of parasite DNA sequence reads relative to host reads. A total of 868,476 genome-wide SNPs were identified, filtered to 104,583 SNPs after exclusion of the hypervariable subtelomeric regions. Phylogenetic analysis showed a clear separation of isolates sourced from Africa and Asia, similar to observations from the analysis of sequence data from the circumsporozoite (pmcsp) gene [148]. Many non-synonymous SNPs in orthologs of P. falciparum drug resistance-associated loci (pmdhfr, pmdhps and pmmdr1) were detected [21,52], but their impact on drug efficacy remains unknown. Thus, to date, there are no validated molecular markers of drug resistance in P.     malariae parasites although, as noted above, prophylaxis breakthrough, treatment failures and emergence following treatment for other species have been reported [26][27][28][29]149].
In the wider Plasmodium species context, phylogenetic analysis has shown that P. malariae isolates group with malariae-like species that infect monkeys and non-human primates [2,23]. Plasmodium malariae parasites also cluster closer to P. ovale spp., but in separate clades, and more generally in a clade with P. vivax, P. knowlesi and Plasmodium cynomolgi that is distant from the Laverania sub-genus exemplified by  [2,150]. Given the range of primate hosts that are infected by P. malariae, P. brasilianum and their close relatives, further genomic studies are needed to tease out the two main questions raised by the studies so far: o Should P. brasilianum, as is currently circulating in South America, and P. malariae be considered distinct, non-recombining species? o What is the extent of the radiation of P. malariae-like species in the great apes?

History & discovery
First identified in Liverpool by Stephens in 1918, the index case of ovale malaria was a British army private, returning to the UK in 1918 following deployment in "East Africa", and having reported an episode of symptomatic malaria in December, 1916 [151]. This soldier's blood films were examined over several months, with no mention of any treatment being offered, during which time the presence of fimbriated, oval infected red cells was noted as a key feature, together with a 48 h fever periodicity. This "new parasite of man" (sic) was thus characterized as a benign tertian infection and named Plasmodium ovale in the primary paper, published in 1922. Some additional detailed description of the parasite and its presentation was published by Stephens and Owen in 1927 [152]. For much of the twentieth century, ovale malaria remained a minor entrant in parasitology textbooks, including Coatney et al. [1], until the advent of molecular diagnostic studies in the 1990s began to uncover evidence of genetic dimorphism [153], leading to a series of papers in the first decade of the twenty-first century examining the impact of this dimorphism on molecular and antigen-based diagnosis [154][155][156][157][158]. A multi-centre effort to gather 51 geographically diverse parasite isolates and generate sequencing data across seven genetic loci was then able to demonstrate that ovale malaria was the result of infection by either of two non-recombining, sympatric sibling parasite species, which were named P. ovale curtisi and P. ovale wallikeri [159]. In the decade that followed, various molecular tools were developed to distinguish the two ovale species, and there was an explosion of our understanding of the contribution of the newly recognized parasites to malaria burden across the tropics.

Distribution and abundance
Although the original identification of P. ovale sensu lato (s.l.) by Stephens was in a British soldier who contracted malaria in "East Africa", the species was subsequently recognized as highly endemic in West Africa (especially Nigeria). Coatney et al. described the distribution of the species as extending to the East African Coast, and as far south as Mozambique [1]. Outside Africa, ovale malaria was sporadically reported from Papua New Guinea, Indonesian islands and some South-East Asian countries [144]. However, with the introduction of molecular diagnostic tools and recognition and widespread acceptance of the two sympatric species, P. o. curtisi (former "classic" type) and P. o. wallikeri (former "variant" type) [159], a much more complex understanding of these parasites has developed. Molecular diagnostics have greatly facilitated the confirmation of the presence of ovale malaria parasites in much of Africa and Asia, including countries where it was not previously known to be present (e.g. Bangladesh, Afghanistan, Angola) [35][36][37][160][161][162], and in non-human primates [163]. However, it remains generally accepted that these parasites are not endemic in the Americas [159].
Infections with ovale malaria parasites are often asymptomatic and parasite densities low, leading to difficulties in accurate microscopic diagnosis and some uncertainties as to distribution in the recent past. Given the presence of intra-erythrocytic stippling on thin films, and the irregular shapes adopted by ovaleinfected cells, there is some morphological similarity to P. vivax, which exacerbates diagnostic difficulties. This also influenced early phylogenetic thinking; Coatney and colleagues write that "from the vivax-like stem developed a morphologically similar species, P. ovale, that was capable of surviving in (African) hominids …" (1). Moreover, mixed infections with other human malaria parasites are very common. Double infections of P. ovale curtisi and P. ovale wallikeri in the same individual have also been reported (e.g. Angola, Bangladesh) [36,161], confirming the lack of recombination between the two species. However, reported prevalence estimates vary widely among various studies, reflecting different study designs and blood sample collection strategies (e.g. asymptomatic vs. febrile patients). The known distribution of P. ovale spp., P. o. wallikeri and P. o. curtisi is presented in Fig. 2, and a detailed listing of reports identifying these species, including GenBank accession ID where relevant, is given in Table 3 [27,36,48,58,72,76,83,90,97,102,106,116,118,137,156,159,.

Genomic studies of P. o. curtisi and P. o. wallikeri
In the period since the two genetically distinct forms of P. ovale spp. were recognized, there have been a limited number of studies that have explored the differences between them. A study in UK travellers with ovale malaria by Nolder and colleagues could not identify any robust features of morphology that can distinguish P. o. curtisi from P. o. wallikeri [168], but were able to provide evidence of a significant difference in the distribution of relapse periodicity: the former species displayed a geometric mean latency of 85.7 days (95% CI 66.1 to 111.1, N = 74), compared to the significantly shorter 40.6 days (95% CI 28.9 to 57.0, N = 60) of the latter. This contrasts with the earlier observation of Chin and Coatney, who conducted studies of experimentally infected volunteers whose initial infections (all with the same "West African strain") were treated with quinine or chloroquine before extended follow-up for evidence of P. vivax-type relapse [218]. These authors concluded that "These results leave little doubt that ovale malaria is a relapsing disease, but there appears to be no definite relapse pattern…" Subsequent studies in European travellers, a group in which super-infection is absent as a potential confounder, have confirmed this difference in latency period between P. ovale curtisi and P. ovale wallikeri [168,219,220]. These studies were also consistent in finding that P. ovale wallikeri is associated with low platelet counts and thus more likely to elicit clinical thrombocytopenia, and more likely to be correctly identified by immunochromatographic lateral flow tests that detect the LDH antigen, which fail to identify > 90% of P. ovale curtisi infections, a reflection of differences in the amino acid sequence of LDH in the two species [158,159].
Given the absence of distinguishing morphological characters, despite reliable differences in some clinical and diagnostic features, there has been increasing attention to characterisation of the genomic organisation of the two sibling species as a route to better understanding their divergence from each other, and to describe the level of within-species diversity. Initial efforts were based on direct sequencing of PCR-amplified loci, and gave a general picture of fixed differences in both synonymous and non-synonymous substitutions between the species in almost every coding region examined, but very little intra-species genetic diversity [159-161, 185, 210, 211]. This was also true of genes related to sexual stage development, which had been examined for evidence of

Poc: Plasmodium ovale curtisi
Pow: Plasmodium ovale wallikeri       [163] a mating barrier between the two species [181]. Whole genome analysis would clearly be very informative, but very few draft genomes of either species are available due to the difficulty in obtaining parasite DNA from these typically very low parasitaemia infections. The first partial genomes to become available were assembled from Illumina short-read sequencing of two isolates of P. o. wallikeri from Chinese workers returning from West Africa, as well as one P. o. curtisi isolate also from a Chinese worker returning from West Africa and the genome of the chimpanzee-propagated Nigeria I strain [1,22,24]. Subsequently, three partial genomes of P. o. curtisi from two patients that tested positive for P. falciparum in Ghana and one mixed infection from Cameroon, together with two P. o. wallikeri genomes obtained from individual patients in Cameroon, were also assembled [23]. Analysis of the P. ovale spp. genomes published to date has estimated a total genome length for both species of ~ 35 Mbp (29% GC content), with 40% being subtelomeric [22,23]. Differences in total length (maximum observed 38Mbp) were observed between isolates, primarily due to differences in the estimated size of expansion of the ocir/owir gene families. These species have considerably more pir genes (1500-2000), than other human plasmodium parasites (~ 300) [25]. A larger number of surfin genes have also been identified, with > 50 present in P. o. curtisi and > 125 in P. o. wallikeri. The variant protein isoforms expressed by members of these gene families may be important for interactions with multiple host ligands and, as they are likely to be antigenically variant, their expansion is thought to have been driven by host immune pressure. Expansion of reticulocyte binding-like proteins (RBP), involved in red blood cell invasion, has been observed in both ovale genomes (13-14 genes), gene copy numbers similar to P. vivax, while in other species only ~ 2-8 copies have been identified. An expansion of the Plasmodium ookinete surface protein P28 appears to be a specific feature of both P. ovale spp, as only one copy appears to exist in the genomes of other human-infecting species in the genus.
All the available data confirm that there is a close genetic relationship between the two species, supported by phylogenetic analysis that show P. o. curtisi and P. o. wallikeri grouping together in the same clade in all studies to date [2,23,159]. However, many differences between the two taxa have been observed when comparing surfin, pir and rbp genes, as isoforms with identical sequences have been observed between isolates of the same species, but these families are far more divergent in between-species comparisons of the few P. o. curtisi and P. o. wallikeri genomes assembled so far. Significant dimorphism has previously been reported in candidate genes across larger datasets from Asian and African isolates [159-161, 175, 185, 210, 211]. For example, specific analysis of nucleotide sequences of five protein-coding regions, likely involved in life cycle sexual stages and so potentially contributing to mating barriers, found that intra-species variation was minimal at each locus, but clear dimorphism were detected when comparing P. o. curtisi to P. o. wallikeri [181]. Similar results were observed across three vaccine candidate surface proteins in samples collected from Thailand and countries in Africa [185], and in multi-locus sequence analyses reported in a large study of both species in Bangladesh [161]. To better understand the intra-and inter -genetic diversity of these species, more complete reference genomes are needed, as well as a much greater number of isolates undergoing whole genome sequencing across geographic regions.

Likely origin of these two closely-related, sympatric and non-recombining species
The question as to how two non-recombining sibling species have ended up co-circulating in the same mammalian hosts, transmitted by the same arthropod vectors, has attracted some attention, as has the difficulty in estimating when the two lineages diverged, and in which primate hosts [2,3,23,25,159]. A thorough summary of the current thinking can be found therein, but the most parsimonious explanation for the current co-circulation of P. o. curtisi and P. o. wallikeri, in what appears to be perfect sympatry, can be paraphrased from reference 26: pre-ovale parasites in an unknown non-human primate host underwent an initial host transition into hominids some millions of years before the present. This new lineage thus began from a single event, representing an extreme genetic bottleneck, and developed apart from the progenitor stock. Substantial genetic drift occurred, while the two parasite lineages were partitioned in different hosts, a form of allopatry. When a second transition into hominid hosts occurred, again through an extreme genetic bottleneck, both lineages now shared the same hosts, but there was insufficient genetic similarity for fertilisation, meiotic pairing and recombination to occur. However, as the two new species shared almost all features of biology and life history, they together flourished in settings where conditions were favourable and appropriate vectors abundant, and both perished where conditions were harsh. This provides a plausible scenario to explain the contemporary observation that P. o. curtisi and P. o. wallikeri are now always found co-circulating in the same host and vector populations. Considering these observations, and the irrefutable evidence assembled since 2010 that the ovale parasites represent two distinct sibling species, it is clear that the trinomial nomenclature currently in use is not fit for purpose. Some of the arguments around this can be found in Box 2 of reference 26; to resolve this situation, the current authors and collaborators have developed a proposed solution in which two new binomials are utilized in place of the current nomenclature (manuscript in preparation). In the meantime, correspondence on this topic is most welcome.
As to the evolutionary origins of the ovale parasites, despite twentieth century phylogenetic analyses in general favouring kinship with P. vivax [1,221], genomic sequencing and elucidation of nuclear protein-coding, ribosomal RNA-coding, and mitochondrial genes have more recently placed these species distant from the vivax clade, which includes P. cynomolgi, P. knowlesi and other SE Asian parasites of simian hosts. Rather a position closer to P. malariae [159], Lemuroidea [222], or perhaps the rodent parasite clade [23], have also been put forward. As more genomic information becomes available for P. o. curtisi and P. o. wallikeri the kinship of these species, and therefore identification of their closest contemporary relatives, should become clearer.

Concluding remarks
Multi-population genomic studies of the neglected malaria parasites considered here are essential to provide insights into the biology underlying mechanisms of infection, disease progression and adaptation to different hosts. Many questions, for these and other Plasmodium species, remain answered, including the ability of some species to form dormant stages in the liver (hypnozoites) as observed for P. vivax and P. ovale species, and suggested as also possible for P. malariae [26], and the regulation of the blood stage cycles that can differ among species (e.g., P. malariae has a quartan cycle, a quotidian cycle is observed for P. knowlesi, while the other primate species all follow a tertian cycle).
Although genomics studies of these parasites have been difficult, the development of new assays such as SWGA allow the whole genome sequencing of parasite DNA from clinical samples [21], and have therefore opened up new opportunities to understand genomic diversity.
Sequencing developments, such as real-time selective sequencing using Nanopore technology, will favour the selection of parasite DNA molecules for sequencing while excluding human molecules [223]. Phenotypic studies of important characters such as drug susceptibility are challenging for these species [224], but the recently developed strategy of "orthologue exchange" now permits detailed in vitro studies of gene function for every species, using transgenic lines with P. falciparum or P. knowlesi as the recipient parasite cell. These and future advances can support the large-scale and cost-effective genomic studies of neglected malaria that are now needed. The resulting gains in knowledge will greatly assist the design of species-specific diagnostics, treatments, and surveillance tools, thereby supporting malaria elimination goals.