Ecological and evolutionary patterns in the enigmatic protist genus Percolomonas (Heterolobosea; Discoba) from diverse habitats

The heterotrophic flagellate Percolomonas cosmopolitus (Heterolobosea) is often observed in saline habitats worldwide, from coastal waters to saturated brines. However, only two cultures assigned to this morphospecies have been examined using molecular methods, and their 18S rRNA gene sequences are extremely different. Further the salinity tolerances of individual strains are unknown. Thus, our knowledge on the autecology and diversity in this morphospecies is deficient. Here, we report 18S rRNA gene data on seven strains similar to P. cosmopolitus from seven geographically remote locations (New Zealand, Kenya, Korea, Poland, Russia, Spain, and the USA) with sample salinities ranging from 4‰ to 280‰, and compare morphology and salinity tolerance of the nine available strains. Percolomonas cosmopolitus-like strains show few-to-no consistent morphological differences, and form six clades separated by often extremely large 18S rRNA gene divergences (up to 42.4%). Some strains grow best at salinities from 75 to 125‰ and represent halophiles. All but one of these belong to two geographically heterogeneous clusters that form a robust monophyletic group in phylogenetic trees; this likely represents an ecologically specialized subclade of halophiles. Our results suggest that P. cosmopolitus is a cluster of several cryptic species (at least), which are unlikely to be distinguished by geography. Interestingly, the 9 Percolomonas strains formed a clade in 18S rRNA gene phylogenies, unlike most previous analyses based on two sequences.


Light microscopy
Live flagellates mounted on glass slides were observed with phase contrast microscopy or differential interference microscopy using a Leica DM5500B microscope equipped with a DFC550 digital camera (Leica, Wetzlar, Germany) or Carl Zeiss AxioScope A.1 microscope equipped with a AVT HORN MC-1009/S analog video camera. To observe the number and shape of the flagella, cultures (1 ml) were centrifuged at ×2,000 g for 10 min, then 900 μL of the supernatant was discarded, and the remaining volumes (i.e. 100 μL) were fixed by addition of 50 μL of 25% v/v glutaraldehyde (electron microscopy grade). The sizes of the live cells (i.e. 50 cells per culture) were measured from digital images. One-way analysis of variance (ANOVA) was used to evaluate variation among strains in their major cell dimensions, using SPSS for windows (Version 25, SPSS).

Scanning electron microscopy
Cultures (1 ml) were centrifuged at ×1,200 g for 10 min, then 900 μL of the supernatant was discarded, and the remaining volumes (i.e. 100 μL) were fixed by adding 50 μL of 25% v/v glutaraldehyde (electron microscopy grade). Fixed cells were allowed to settle (40 min) on glass coverslips coated with 1% poly-L-lysine. Cells were rinsed with sterile media, and then dehydrated with a graded ethanol series (30-100%). The glass coverslips were then critical-point dried. Fixed cells were coated with gold/platinum using an ion sputter system. Specimens were examined with a SU8220 field emission scanning electron microscope (Hitachi, Tokyo, Japan) or JSM-6510LV scanning electron microscope (JEOL Ltd., Tokyo, Japan).

Molecular sequencing and phylogenetic analysis
Nucleic acids from the six new isolates and P. lacustris HLM-6 (i.e. total seven strains) were extracted using a DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) or MasterPure Complete DNA and RNA Purification Kit (Epicentre, Madison, USA), as described in the supplied protocols. For all strains except LRS and HLM-6, the 18S rRNA gene sequences were obtained by PCR amplification using a combination of the eukaryote primers EukA 5 0 -AACCTGGTTGATCCT GCCAGT-3 0 and EukB 5 0 -TGATCCTTCTGCAGGTTCACCTAC-3 0 [40]. The 20-μL PCR reactions included 1.5 μL each of 10-μM stocks of the primers, 2 μL of a 0.25-mM dNTP-mix, 0.8 μL of 50 mM MgCl 2 , 0.2 μL of 5 U/μL Taq DNA polymerase (Solgent, Daegeon, Republic of Korea), and 1-3 μL of DNA template. The cycling conditions were as follows: an initial denaturing step at 94˚C for 5 min, followed by 35 cycles of 30 s at 94˚C, 1 min of annealing at 55˚C, and extension at 72˚C for 2 min, with a final extension step for 10 min at 72˚C. Amplicons were cloned into a pGEM-T Easy vector, at least five positive clones per sample were partially sequenced, and a positive clone was completely sequenced using various sequencing primers. For strain XLG1-P, the cycling conditions were slightly different: 35 cycles of 20 s at 94˚C, 1 min of annealing at 55˚C, and 3 min of extension at 72˚C. Strain LRS was amplified using the different eukaryote primers 82F (5'-GAAACTGCGAATGGCTC-3') and 1498R (5'-CACCTACGGAAACCTTGTTA-3'). The optimized PCR condition was 2 min at 96˚C (an initial denaturation), followed by 35 cycles of 30 s at 96˚C, 1 min at 60˚C, 2 min at 72˚C, with a final extension for 10 min at 72˚C. The PCR products for strains XLG1-P and LRS were directly sequenced by Sanger dideoxy sequencing without cloning.
The 18S rRNA gene sequences from 62 representative heterolobosean species, plus 16 other Discoba species selected as outgroups, were used for phylogenetic analysis (the seed alignment originated from Jhin and Park [33]). The dataset was aligned and masked by eye, with 1,033 unambiguously aligned sites retained for analysis. The alignment is available on request. Phylogenetic trees were inferred by Maximum Likelihood (ML) and Bayesian analyses. The GTR + gamma + I model of sequence evolution was selected for the dataset using MrModeltest 2.2 [43] and was used for both analyses. The ML tree was estimated using RAxML-VI-HPC v.7 [44] with the GTRGAMMAI model setting, 500 random starting taxon addition sequences, and statistical support assessed using bootstrapping with 10,000 replicates. The Bayesian analysis was conducted in MrBayes 3.2 [45] with two independent runs, each with four chains running for 2 × 10 7 generations with the default heating parameter (0.1) and sampling frequency (0.01). A burn-in of 30% was used, by which point convergence had been achieved (the average standard deviation of split frequencies for the last 75% of generations was < 0.05).

Salinity ranges for growth
To estimate the salinity ranges supporting growth of the six new isolates and three previously available isolates, we performed an experiment using media with 3‰ to 300‰ salinity, made from artificial seawater stock (Medium V; see above) as described earlier [33,34,36,37,46]. In brief, the medium was supplemented with heat-killed Enterobacter aerogenes at an initial density of 3.46 × 10 7 cells per ml (20 μL) at 7-to 14-day intervals to support the growth of the protists. All treatments were performed in duplicate. Medium V (0.96 ml) with a range of salinities (3‰-300‰) were inoculated with 20 μL of actively growing stock culture (100‰ or 35 ‰ salinity media with autoclaved barley grain) and incubated in the dark at 25˚C for at least 49 days. We confirmed the salinity range supporting growth by transferring a sample of the isolate into fresh media with the same salinity (0.96 ml of media; inoculum size 20 μL), and re-examining the culture for actively moving cells at 7-to 14-day intervals over a period of 35 days.

General morphology
Live cells were usually ovoid-shaped or spindle-shaped with average lengths and widths ranging from 4.6 to 9.4 μm and 2.9 to 4.3 μm, respectively (Fig 1 and Fig 2). The ratio of length and width of the cells was between 1.8 and 3.3 on average (Fig 1). The biggest cells (length: 9.4 μm, width: 4.3 μm on average) were from strain LO isolated from Lake Turkana, Kenya, whereas the smallest cells were from strain ATCC 50343 (length: 4.6 μm, width: 3.0 μm on average), but the sizes of different strains represented an overlapping continuum. Cells had four flagella inserted in the sub-anterior part of the cell, at the head of the ventral cytostomal groove (Fig 2  and Fig 3). Three flagella were shorter, and one flagellum was longer. The three short flagella were similar in length, similar to that of the cytostomal groove (typically~4 μm long). The long flagellum averaged 15.0-18.9 μm in length, depending on the strain, which was 1.8-3.3 times the length of the cell body (Fig 1). Most cells had an acroneme at the tip of the long flagellum (Fig 3), however, this seems not to be a permanent feature. All species showed jerking motility and sometimes rotated in a counterclockwise direction. The left side of the cytostomal groove was curved, whereas the right side was roughly linear (Fig 3). No amoeboid form was observed during the cultivation of the organisms, and a cyst form was observed occasionally in all cultures except for strain ATCC 50343. Mostly, we did not observe discrete features that distinguished isolates from each other using light or scanning electron microscopy. The length of the long flagellum in the previously described P. lacustris (strain HLM-6) was not significantly different from other strains, except strain ATCC 50343 (ANOVA, p<0.01). Strain HLM-6 had significantly more elongated cells than ATCC 50343 (ANOVA, p<0.01), but the width was not significantly different (data not shown). Probably, morphological dimensions studied here may not be a reliable criterion for Percolomonas taxonomy.

Molecular phylogeny of 18S rRNA gene sequences
The 18S rRNA gene sequences from the seven Percolomonas strains were closest by BLASTN search to Percolomonas cosmopolitus strain WS (AF 519443), but with a low identity of 74% to 83%. Heterolobosea formed a strong monophyletic group in phylogenetic trees of 18S rRNA gene sequences (Fig 4). The seven new sequences and two previously published sequences from Percolomonas strains branched with the pseudociliate taxon Stephanopogonidae with strong support (100% ML; PP 1; Fig 4) forming the clade Percolatea. Interestingly, Percolomonadidae, including the seven new sequences, formed a monophyletic group, with moderate bootstrap support (ML: 76%) and posterior probability 1. Within Percolomonadidae, ATCC 50343 (USA) and P5-P (Poland) were distinct from each other and the remaining seven strains, which formed a maximally supported clade. These subdivided further into (i) WS (Russia), (ii) HLM-6 (Russia), and (iii) a maximally supported 'putative halophile clade' of 5 strains in 2 clusters, 'A' and 'B', for a grand total of 6 distinct clusters or single-strain lineages. The 18S rRNA gene sequences differences between these 6 clusters range from 15.1 to 42.4%. Group A of the putative halophile clade was composed of strains LRS (Spain), XLG1-P (New Zealand), and S4 (Korea) and formed a maximally supported clade (Fig 3), with low genetic divergence among strains (98% to 99% identities). All members were isolated from hypersaline waters of 180 to 280‰ salinity (Table 1). Group B consisted of strains LO (Kenya) and SD2A (USA) which showed 98% sequence identity, though formed a weakly supported clade (62% ML; PP 0.75). Strain SD2A was isolated at 200‰ salinity, while strain LO was isolated from a low salinity (4‰) sample, though it cannot grow at this salinity (see below and Table 1).

Morphology of Percolomonas isolates
In general, the isolates studied here are morphologically indistinguishable by light and scanning electron microscopy from the seminal modern accounts of Percolomonas cosmopolitus https://doi.org/10.1371/journal.pone.0216188.g001 [25,26]. They all have four flagella, specifically three short flagella (~4 μm long) and one long flagellum (12-20 μm long) directed along a ventral groove. The size of the cell body of P. cosmopolitus is 6-12 μm in length, 3-9 μm in width [25], overlapping with all of our isolates. The curved left side of the cytostomal groove differs from the right side with a near-linear shape. All strains move with a jerky gliding motion. Thus, it appears that all new isolates may be assigned to the morphospecies P. cosmopolitus. The nominal Percolomonas species most similar to P. cosmopolitus include P. denhami, with three flagella, two of which are long [47] and P. similis with two flagella, one short, and the other long [48]. There are no molecular data for either of these species. The third similar species is Percolomonas lacustris, which was introduced based on the description of strain HLM-6. This was distinguished morphologically from P. cosmopolitus primarily by having a small posterior bulbous projection and an acroneme on one short flagellum [38]. However, the bulbous projection appears by light microscopy just as a pointed tip, which was a feature of the original description of P. cosmopolitus [24], and was seen intermittently in several of our isolates (see Fig 2). We confirmed that one Halophile clade short flagellum bearing a proportionately long acroneme can be a feature of strain HLM-6 (compare our Fig 2F to Fig 2G in [38]), but note that other cells of the strain appear to lack it (Fig 2F in [38]), and that (shorter) acronemes can be observed on the short flagella of several of the strains studied here (our Fig 3). Thus, it seems that the morphological distinction between P. lacustris and other P. cosmopolitus-like strains is subtle at best.

Cryptic species complex in Percolomonas
The Percolomonas complex includes 6 distinct evolutionary lineages that are separated by genetic distances of 15.1% or more. Considering the similar morphology of P. cosmopolituslike strains (see above), it is reasonable to regard the Percolomonas complex as a grouping of several (at least) cryptic (or sibling) species. Among protists, it has long been recognized that numerous cryptic species can exist within a single morphospecies [4,23,[49][50][51][52]. Fenchel [50] suggested that the cryptic protists are regarded as allopatric speciation, and are distributed in limited geographic regions. In our case, however, no geographic clustering of Percolomonas strains was observed in the two clades represented by multiple isolates, although the number of sampling sites and sequencing data are limited. Three Percolomonas strains in 'Group A' form a robust clade with low genetic divergences (98% to 99% identities) and are widely dispersed geographically (Korea, New Zealand, and Spain). In addition, two Percolomonas strains in 'Group B' (with 98% identity) are from Africa and North America. This pattern is more consistent with widespread dispersal, as inferred for other very small protists (e.g. [53][54][55]), and we speculate that environmental selection to different habitats may determine which groups appear where (see below).
In principle, subgroups of a single morphospecies could be differentiated by subcellular structures, and/or minute ultrastructural distinctions (i.e. represent pseudocryptic species [56,57]). It is possible that detailed ultrastructural observations of Percolomonas strains (beyond the scope of this work) may reveal some fixed, morphological differences. There may also be differences in the cyst stages (P. lacustris strain HLM-6 possesses one cyst pore with a plug, but comparable data are lacking for other strains, and unfortunately the ability to make cysts can be lost in culture; [25]). If so, it would be interesting to examine whether such differences distinguish one or more of the six phylogenetic groupings we identified, or instead delineate larger or smaller clades.

Halophily in Percolomonas
Ruinen [24] reported that P. cosmopolitus was detected at salinities ranging from 30‰ to saturated brines. In our study, the nine examined Percolomonas strains derived from a wide range of salinities, up to near-saturation. In general, the degree of halophily of these strains was related to the salt concentration of their original habitats. Five Percolomonas strains (i.e. LRS, SD2A, XLG1-P, S4, and P5-P) were isolated from various habitats with a salinity range of 73‰ to 280‰. These all grew best at 75‰-125‰ by our qualitative estimation and most could still grow at 175‰ or even 200‰ on our experimental media. This clearly makes them halophiles according to the definition by Oren [58], in which a halophile could grow at 50‰ or higher and tolerated at 100‰ salinity. None were obligate halophiles, however, since all could also grow at 30‰ salinity. By contrast, three Percolomonas strains (i.e. HLM-6, ATCC 50343, and WS) isolated from non-hypersaline habitats (20‰ to 36‰ salinity) grew optimally at 30‰ or 50‰, and failed to grow above 100‰ salinity. Interestingly, Percolmonas strain LO grew optimally at 50‰ salinity, and grew up to 100‰ salinity, although it was isolated from Lake Turkana, Kenya, with a very low salinity of 4‰. Phylogenetically, LO belongs to the putative 'halophile clade', and it is possible that it descended from a more halophilic ancestor (see below). Interestingly, the Lake Turkana sample was also the origin for Pharyngomonas turkanaensis, a heterolobosean amoeba that grows best at 15-30‰ salinity, but is inferred to have descended from a halophilic Pharyngomonas ancestor [59].
The taxon Heterolobosea includes a substantial proportion of the known halophilic or halotolerant eukaryotes [34][35][36][37][59][60][61][62], and is interesting for examining the evolution of halophiles [33,60,61]. Recently, Kirby et al. [61] and Jhin and Park [33] suggested that the Tulamoebidae clade (sensu lato) in Heterolobosea were an example of a radiation of morphospecies that stemmed from a common halophilic ancestor. This clade included Pleurostomum flabellatum, Tulamoeba peronaphora, Tulamoeba bucina, and Aurem hypersalina, all with optimal salinities for growth of at least 150‰ [33,36,37,61]. It is possible that the Percolomonas clade consisting of 'Group A' and 'Group B' may also represent a radiation of halophiles (albeit one of cryptic species within a morphospecies). All of the cultivated Percolomonas in this clade are halophiles, with one borderline case (LO; see above). However, this possibility of an exclusively/predominantly halophile clade could be a biased sampling artefact, and should be tested through additional isolations and study of related strains. This would also be useful to understand the nature of the closest relative of the halophile strain P5-P, which is phylogenetically isolated from others at present.

Is Percolomonas monophyletic?
For a long time only two 18S rRNA gene sequences of P. cosmopolitus were available, and these were included in many phylogenetic analyses of Heterolobosea [28,[30][31][32][33][34][35][36][37][63][64][65][66]. Most of these phylogenies showed the two sequences forming a paraphyletic group, with one more closely related to Stephanopogon [65,66]. This inference may have been affected by the small number of sequences of Percolomonas available. In the present study, with seven additional and different 18S rRNA gene sequences, we instead inferred a (moderately supported) Percolomonas clade. Future research will address the cause of this difference, and test the relationships amongst Percolomonas and Stephanopogon using other markers.

Conclusions
On the basis of light and scanning electron microscopic observations, all Percolomonas strains studied here are morphologically very similar, in spite of the huge genetic diversity they encompass. Percolomonas strains form at least 6 genetically distinct clades in the molecular phylogenetic trees, and could be considered to represent at least as many cryptic species. The speciation of Percolomonas strains could be partially related to salinity preference, rather than spatial distribution. The clusters 'Group A' and 'Group B', which are specifically related, may collectively represent a halophilic clade.