Integrating coalescent species delimitation with analysis of host specificity reveals extensive cryptic diversity despite minimal mitochondrial divergence in the malaria parasite genus Leucocytozoon

Abstract

Background: Coalescent methods that use multi-locus sequence data are powerful tools for identifying putatively reproductively isolated lineages, though this approach has rarely been used for the study of microbial groups that are likely to harbor many unrecognized species. Among microbial symbionts, integrating genetic species delimitation methods with trait data that could indicate reproductive isolation, such as host specificity data, has rarely been used despite its potential to inform species limits. Here we test the ability of an integrative approach combining genetic and host specificity data to delimit species within the avian malaria parasite genus Leucocytozoon in central Alaska.

Results: We sequenced seven nuclear loci for 69 Leucocytozoon samples and used multiple species delimitation methods (GMYC and BPP models), tested for differences in host infection patterns among putative species based on 406 individual infections, and characterized parasite morphology. We found that cryptic morphology has masked a highly diverse Leucocytozoon assemblage, with most species delimitation methods recovering support for at least 21 separate species that occur sympatrically and have divergent host infection patterns. Reproductive isolation among putative species appears to have evolved despite low mtDNA divergence, and in one instance two Leucocytozoon cytb haplotypes that differed by a single base pair (~ 0.2% divergence) were supported as separate species. However, there was no consistent association between mtDNA divergence and species limits. Among cytb haplotypes that differed by one to three base pairs we observed idiosyncratic patterns of nuclear and ecological divergence, with cytb haplotype pairs found to be either conspecific, reproductively isolated with no divergence in host specificity, or reproductively isolated with divergent patterns of host specialization.

Conclusion: Integrating multi-locus genetic species delimitation methods and non-traditional ecological data types such as host specificity provide a novel view of the diversity of avian malaria parasites that has been missed previously using morphology and mtDNA barcodes. Species delimitation methods show that Leucocytozoon is highly species-rich in Alaska, and the genus is likely to harbor extraordinary species-level diversity worldwide. Integrating genetic and ecological data will be an important approach for understanding the diversity and evolutionary history of microbial symbionts moving forward.

Keywords: Avian malaria; BPP; Cryptic species; GMYC; Haemosporida; UniFrac.

Fig. 1

Summary of species delimitation results. Shown is a seven-gene phylogeny estimated in BEAST using the partitioned nuclear sequence, a strict molecular clock, and coalescent prior. To the left of the phylogeny are images of the three morphospecies found in this study and a map of the sites where all samples were collected generated using the ggmap package in R with the “get_googlemap()” function. Columns to the right of the phylogeny are the morphospecies for each sample (where available), and the results of four species delimitation methods: the A11 algorithm of BPP, single-threshold GMYC, bGMYC, and multiple-threshold GMYC. If multiple samples were supported by a species delimitation method to represent the same species, that putative species is represented in a unique color in a rectangle separated from other species. Haplotype names for each sample (or group of samples) are shown; where species delimitation methods recovered multiple haplotypes as the same species, the most common haplotype of the group is given first. Note that haplotypes CATUST09 and TUMIG15 were not included in the BPP analyses and so are not depicted in the column for BPP

Fig. 2

Nuclear divergence of Leucocytozoon cytb haplotypes. Depicted are five clades that contain pairs of cytb haplotypes that differ by one to three base pairs (as shown in the distance matrices of cytb differences to the right of the networks). In each clade haplotypes are given a different color that corresponds to the haplotype networks. a haplotypes PERCAN01, ACAFLA03, and CATUST14, which differ by one to three bases, are differentiated across all nuclear loci and never share nuclear alleles. The three sequences that are marked by asterisks represent outlier samples of high divergence that were removed prior to species delimitation analysis due to the possibility that they contained mixed infections; b) ROFI06 and LOXLEU05 differ by a single base pair and are weakly differentiated across nuclear loci, sharing nuclear alleles for a single locus; c) ZOLEU02 and ZOLEU04 differ by a single base pair and share alleles across multiple nuclear loci; d) haplotypes CB1, JUNHYE04, CARPUS01, and ACAFLA04 differ by one to three bases, but share alleles across multiple nuclear loci. In addition, haplotype pairs TRPIP2/ACAFLA01 and BT1/PHYBOR01 differ by a single base pair but share nuclear alleles; e) CATUST28 and CATMIN05 differ by two bases but are highly differentiated across all nuclear loci. Note that not all nuclear loci are depicted for each clade due to insufficient sample sizes for some haplotypes

Fig. 3

Results of BPP species delimitation using the A11 algorithm. Boxes at each node depict the results of the four analyses that were conducted using different diffuse priors for population size and divergence time. If a node was supported to delimit a species with posterior probability > 0.95, the square for that analysis is depicted in blue (squares are depicted in yellow and red for posterior probabilities less than 0.95 and 0.90, respectively). Note that no posterior probabilities are depicted for lineages CATUST09 and TUMIG15, as they were removed from this analysis due to high divergence from all other haplotypes

Fig. 4

The relationship between cytb and nuclear divergence in Leucocytozoon is inconsistent and clade-dependent. a Among 14 pairs of Leucocytozoon cytb haplotypes (17 haplotypes total), the divergence between them is not strongly associated with nuclear divergence for the seven loci sequenced in this study. b The relationship between cytb divergence and nuclear divergence appears to be clade dependent, as the clade consisting of CB1/ACAFLA04/CARPUS01/JUNHYE04 (which were all found to be conspecific) exhibits no association between cytb divergence and nuclear divergence while the clade consisting of ACAFLA03/CATUST14/PERCAN01 exhibits strong nuclear differentiation across the same scale of cytb divergence, though nuclear divergence does not appear to increase with cytb divergence

Fig. 5

Patterns of putative Leucocytozoon species infection across host species. a Principal coordinates analysis of weighted UniFrac distances among 18 Leucocytozoon cytb haplotypes that were sampled at least twice across 381 hosts. b Paired histograms of randomized weighted UniFrac distance and barplots demonstrating differential infection across potential host families for closely related cytb haplotype pairs. Vertical lines in histograms depict the observed UniFrac distance for that putative species pair, which is also shown above each histogram with its associated significance value. Colors of the barplots correspond to the host phylogeny shown in the key to the left. UniFrac distances and barplots are depicted in two pairwise matrices for clades that contain three putative species; UniFrac distance is shown below the diagonal for each cytb haplotype pair, while above the diagonal are host infection barplots with the haplotype along the lower diagonal representing the top barplot