Introduction

Microbial associations are exceptionally common and widespread [1,2,3,4], and those between eukaryotic hosts and the prokaryotic organisms they harbor are the most intensely studied. Historically, the focus has been put on “symbioses”, broadly intended as relationships involving protracted physical contact between the partners [5, 6], regardless of the consequences of the relationship (e.g., mutualism, commensalism, or parasitism). Attention has now shifted to the study not just of one or a few microbial symbionts, but of the entire prokaryotic communities associated with eukaryotic hosts—the “microbiomes” [7]—which include both stable and temporary members (from “true symbionts” to food organisms). Almost invariably, the investigated hosts are large metazoans or plants. And yet, symbioses between prokaryotes and unicellular microbial eukaryotes (protists) are just as common [e.g., 1, 8–12]. Associations between bacteria and ciliates (phylum Ciliophora), in particular, boast a long history of studies which dates back to the nineteenth century [13, 14] and has significantly grown during the last decades [e.g., 15–21]. Even so, the bulk of the literature deals only with a few ciliate host species and genera, such as Paramecium, Euplotes, Spirostomum, and Metopus [e.g., 22– 25]. Among more than 1500 described ciliate genera [26], only about 30 have been screened for the presence of prokaryotic symbionts [27]. A molecular characterization of the associated bacteria has been achieved even less frequently. In the few cases when considerable data exist for a ciliate-harbored prokaryotic symbiont, the focus is usually on a single bacterial species. However, it is known that complex consortia can be found associated with each individual ciliate host [28, 29], true microbiomes within microbes. While detection and characterization of microbiomes associated with multicellular eukaryotic hosts have considerably advanced in the last decades [3, 4, 30, 31], the same topic is almost completely unexplored for unicellular eukaryotic hosts.

Ultimately, the source of almost all these limitations is the scarcity and unreliability of cultivation-independent methods. Ciliate-prokaryote associations have been studied so far mainly in hosts that are easy to maintain in standard laboratory conditions. Among ciliate taxa found in any natural sample, few if any can be reliably established as monoclonal cultures with the stability and abundance required by the full-cycle rRNA approach [32]; moreover, those few do not necessarily represent well the original community. It is therefore reasonable to assume that a huge amount of yet-undiscovered microbial associations, involving uncultivable or difficult-to-cultivate ciliate hosts, is present in the natural environment. The ecological meaning of ciliate-prokaryote associations has been elucidated only in a few instances [1, 27], always using cultivation-dependent approaches [33,34,35,36,37,38]. Field investigations are either completely lacking or performed without a precise identification of the involved partner [39]. Coupling reliable characterizations of the hosts to environmental surveys has been extremely difficult up to now. Hence, an assessment of the distribution and potential adaptive value of ciliate-prokaryote associations in the natural environment is currently impossible. As a consequence, nothing is known concerning the epidemiology of even the most studied symbioses: virtually no quantitative data are available on geographic distributions or habitat preferences, nor on the prevalence of infection in host natural populations. Such basic data are essential to unravel any potential role of these associations in the ecology and evolution of both hosts and symbionts.

A promising approach to address such issues is the application of single-cell “omic” techniques to specimens freshly isolated from samples. Single-cell genomics [40] and transcriptomics [41] have been already successfully performed on ciliates. Prokaryotic gene cloning from single-cell isolates has been occasionally used in the context of field studies [28, 29, 42, 43], but it proved not always sufficient to characterize all the associated bacteria of potential interest [28]. Single-cell ciliate microbiomics, the characterization of prokaryotic communities harbored by a single host cell using high-throughput sequencing techniques, could represent the next step.

Here, we present a pilot single-cell microbiomic study, suggesting that this approach can be easily and efficiently used for field research on microbial associations. Based on ciliate single-cell PCR followed by Illumina sequencing of small subunit (SSU) rRNA gene fragments, the method described does not require host cultivation and allows immediate processing of environmental samples for an accurate characterization of both hosts and associated microbiomes. We tested the approach on ciliates from freshwater and marine environments, of various sizes and belonging to different taxonomic groups. Obtained results encourage its use for rapid and successful surveys of microbial associations between ciliates and prokaryotes in the environment, bringing us closer to fully understand their extent, distribution, and diversity.

Materials and Methods

Sampling and Ciliate Isolation

Sediment-water samples (45 ml each) were collected in the same week, between the end of February and the beginning of March 2017, from two freshwater sandy ponds (P1_1, P3_2) near the mouth of the river Serchio (Pisa, Italy) and from two small rocky tidal pools (L1_4, L2_1) along the Ligurian Sea shore in Livorno (Italy). About 30 ml of medium from each sample were transferred into Petri dishes and observed with a Wild Heerbrugg optical stereo microscope (× 400). As is often the case in freshly collected samples from such habitats [44, 45], ciliate diversity was rather high, while population abundances were low. Twenty-eight ciliate specimens of several morphotypes were individually isolated by glass micropipette (see Table 1). At least one specimen per observed morphotype was processed from each sample: twelve from sample P1_1, ten from sample P3_2, three from sample L1_4, and three from sample L2_1. Isolated ciliate cells were washed to minimize the presence of prokaryotes not tightly attached to the host cell. Freshwater specimens were rinsed one by one three times with sterile mineral water in separate wells, and marine specimens were rinsed with artificial sterile marine water (33‰ salinity). Each single cell was then washed three more times with sterile distilled water and stored in a 0.2-ml tube in 70% (v/v) ethanol. This procedure, followed by storage at − 20 °C, was performed within 48 h from sample collection to reduce the risk of contamination from the lab. The portion of each sample (15 ml) not used to harvest ciliates was fixed in 70% (v/v) ethanol and divided into three aliquots used in our survey as controls, for the purpose of characterizing the background environmental microbial communities.

Table 1 Ciliate identification. NCBI Blastn-based affiliations of isolated ciliate cells based on SSU rRNA genes
Full size table

Eukaryotic and Prokaryotic SSU rRNA Gene Amplification and Sequencing

Ethanol was removed from tubes containing single ciliate cells using a SpeedVac SVC100 (SAVANT). As a first step, simultaneous PCR amplifications of eukaryotic and prokaryotic SSU rRNA genes were performed directly on individually isolated cells in the same tube where each cell was stored (no DNA extraction was performed). The Takara ExTaq (Takara Biochemicals) reaction solution, including primers 18S F9 Euk [46] and 18S R1513 [47] for ciliates, and primers 8F [48] and UNI-b-rev [32] for prokaryotes, was pipetted on top of the cell [45]. In order to identify the hosts, amplicons were purified with the Eurogold Cycle-Pure Kit (Euroclone) and diluted 1:100, then subjected to two semi-nested amplifications, one with eukaryotic primers 18S F9 and 18S R1052 [49], the other with eukaryotic primers 18S F783 [49] and 18S R1513 [47]. Resulting amplicons were further purified and Sanger sequenced using multiple appropriate internal primers [49, 50] by GATC Biotech (Cologne, Germany). In order to characterize the prokaryotic microbiomes of ciliates, a nested PCR was performed on amplicons obtained in the first step. This amplification used the KAPA HiFi HotStart Ready Mix with the prokaryotic primer set for the V3–V4 regions of the SSU rRNA gene suggested by Klindworth and colleagues [51]. The Illumina overhang adapter sequences added to the forward and reverse primers were 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3′ and 5-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3′, respectively (Illumina protocol, Part # 15044223, Rev. B).

In order to characterize background prokaryotic communities, including both free-living and host-associated taxa, total genomic DNA was extracted from 0.25 g of each of the three control aliquots per sample using the PowerSoil DNA Isolation Kit (MoBio). Extracted DNA was used as template for amplification with the KAPA HiFi HotStart Ready Mix and the prokaryotic primer set for the V3–V4 regions of the SSU rRNA gene as described above.

Prokaryotic amplicons from single host cells and controls were barcoded, pooled, and sequenced by BMR Genomics (Padova, Italy) on the Illumina MiSeq platform (2 × 300 paired-end sequencing with MiSeq Reagent Kit v3) by BMR Genomics (Padova, Italy).

Sequence Analysis

Eukaryotic SSU rRNA gene sequences obtained by Sanger sequencing were analyzed with NCBI Blast [52] for putative identification of the ciliate hosts.

Raw reads of prokaryotic V3–V4 regions obtained by Illumina MiSeq were analyzed using the Quantitative Insights Into Microbial Ecology version 2 (QIIME2, https://qiime2.org) software package (v. 2017.2, [53]). Reads were initially truncated at base 290 to remove the lower-quality last 10 base calls. Then, quality filtering, primer trimming, and pair-end read merging were performed with DADA2 [54] (default settings: sequences with any N character discarded; sequences truncated at any base with a quality score of 2 or lower; maximum expected error allowed: 2; chimera removal de novo). Unique reliable sequences (“sequence variants,” each representing one or more identical sequences) were aligned using MAFFT [55], and highly variable positions were masked. A phylogenetic tree was inferred with FastTree [56]. Three long-branching sequence variants were manually inspected and removed as host contaminants (NCBI Blast best hits corresponded to ciliates of the genera Trithigmostoma, Hartmannula, and Sterkiella). A fourth long-branching sequence variant, represented by only two sequences in a single library, was removed because it was clearly chimeric. Taxonomic classification was performed using the Greengenes database [57] release 13.8. Following Werner and colleagues [58], the regions of interest were extracted from SSU rRNA representative sequences (99% similarity clustered Operational Taxonomic Unit) and used to train a Naive Bayes classifier. Sequence variants identified as mitochondria or chloroplasts were removed before further data processing (e.g., bar plots and heatmap building), which was also performed on QIIME2. No remaining sequence was assigned to eukaryotes or eukaryotic organelles.

Statistical Analysis

Before performing statistical analyses, 4569 sequences (the number of merged, quality-filtered reads in the smallest library) were randomly sampled from each library. Four measures of Alpha-diversity were calculated on QIIME2: sequence variant number, Faith’s Phylogenetic Diversity (qualitative index using phylogenetic information), and Shannon’s (quantitative, non-phylogeny-based index) for richness and Pielou’s Evenness for evenness. Rarefaction curves were also inferred. Comparisons among different communities were performed using the Kruskal-Wallis non-parametric test. Beta-diversity analyses were performed using Permanova and multivariate PCoA, testing various metrics: Bray-Curtis and Jaccard for quantitative and qualitative data, respectively, and Uni-Frac distances, both weighted and unweighted, to assess the impact of phylogeny.

Results

Ciliate Hosts Identification

Partial eukaryotic SSU rRNA gene sequences were obtained from each of the 28 single-cell ciliate specimens (minimum length 858 bp, maximum length 1855 bp). The sequences always included the V4 region, recognized as the most variable and information-rich region of the SSU rRNA gene for ciliates [59]. Ciliate identification was then performed using NCBI Blastn (Table 1). Ciliates belonging to 17 different genera, distributed in six different classes [26], were identified. Obtained sequences have been deposited to the ENA database under the accession numbers LT985649-LT985676.

Ciliate Prokaryotic Microbiomes

The final dataset contained 567,919 sequences, with 14,198 ± 7281 mean sequences per library. The library with most reads was obtained from a marine ciliate specimen (L_4, likely belonging to the genus Aspidisca; 40,898 sequences), the one with fewest from a freshwater control (P1_1_1; 4569 sequences). Raw reads have been deposited to the ENA database (study number PRJEB25414). Obtained sequences were grouped in 3575 different sequence variants. The average numbers of sequence variants were 24.3 ± 16.9 (samples P1_1 and P3_2) and 79.3 ± 66.5 (samples L1_4 and L2_1) in freshwater and marine ciliate microbiomes, respectively, and 469.8 ± 127.6 (samples P1_1 and P3_2) and 314.3 ± 146.5 (samples L1_4 and L2_1) for freshwater and marine prokaryotic environmental communities in controls. The difference in richness (number of sequence variants) between ciliate microbiomes and environmental communities (controls) in the same habitat, with microbiomes being considerably poorer, was significant in libraries of marine samples (p = 0.01) and highly significant in libraries of freshwater samples (p < 0.01). Other richness indexes confirmed the pattern, with differences always being highly significant (Online Resource 1). Similarly, evenness was significantly lower in microbiomes than in environmental communities. Differences in richness and evenness measures among controls were not significant or barely significant (p = 0.05).

Four different clusters, i.e., microbiomes associated with freshwater ciliates, microbiomes associated with marine ciliates, communities of freshwater controls, and communities of marine controls, are visible in the arrangement of libraries in PCoA graphs (Fig. 1). The microbiome of the marine ciliate L_1 (likely belonging to the genus Hartmannula) is the only conspicuous outlier, clustering with the microbiomes of freshwater ciliates. The presence in the same library of a single eukaryotic sequence variant (later removed during quality filtering, see the “Materials and Methods” section), identical to the one independently obtained by Sanger sequencing from the Hartmannula host, excludes the possibility of contamination or mislabeling. Permanova tests of the differences between groups are highly significant (p value < 0.01) regardless of the employed metric, with a single exception (the separation of microbiomes from marine ciliates and marine control communities according to the weighted Uni-Frac distances, for which p = 0.011). Results of Permanova tests are reported in Online Resource 2.

Fig. 1
figure 1

Principal component analysis. PCoA graph of prokaryotic communities from controls and single-cell ciliate specimens (unweighted Uni-Frac distances). Similar results were produced using different metrics (weighted Uni-Frac, Bray-Curtis, and Jaccard)

Full size image

Plateaus in rarefaction curves confirm that sequencing depth was sufficient to sample all sequence variants in the libraries (data not shown). In all prokaryotic communities, the most represented phyla were Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes (Online Resource 3). The number of detected phyla mirrors the observed richness trends, being higher in control communities and lower in ciliate microbiomes.

Detection of Potential Prokaryotic Symbionts

Heatmaps of the 100 most abundant prokaryotic taxa (identified at the least inclusive level allowed by the reference database, roughly corresponding to the genus rank) were built for freshwater and marine environments (Fig. 2). In freshwater libraries (Fig. 2a), it is immediately apparent that a group of bacterial taxa abundant in ciliate microbiomes is absent or scarcely represented in controls. Other bacterial genera are well-represented both in freshwater ciliates and controls. Finally, many prokaryotic taxa are abundant in controls, but absent or scarcely represented in ciliate microbiomes. A somewhat similar situation applies to marine libraries, although a larger fraction of the taxa is present and abundant both in ciliate microbiomes and environmental control communities.

Fig. 2
figure 2

Heatmaps of the 100 most abundant freshwater and marine prokaryotic taxa. a Heatmap showing the relative abundances (percentage of sequences) of freshwater prokaryotic taxa identified at the least inclusive taxonomic level in the Greengenes taxonomy. As shown by the trees, rows and columns are arranged according to UPGMA clustering for readability. Bar under the heatmap indicates the three different patterns observed: bacterial taxa abundant in ciliate microbiomes, but absent or scarcely represented in controls (black); bacterial taxa well represented both in freshwater ciliates and controls (pale gray); and bacterial taxa abundant in controls, but absent or scarcely represented in ciliate microbiomes (dark gray). FCM: freshwater ciliate microbiomes. FCC: freshwater control communities. b Heatmap showing the relative abundances (percentage of sequences) of marine prokaryotic taxa identified at the least inclusive taxonomic level in the Greengenes taxonomy. As shown by the trees, rows and columns are arranged according to UPGMA clustering for readability. MCM: marine ciliate microbiomes. MCC: marine control communities

Full size image

The same three patterns can be appreciated in taxa bar plots (Fig. 3). Several well-known prokaryotic symbionts of ciliates are retrieved with high relative abundances (> 5% of the sequences) among taxa detected in ciliate microbiomes. For example, members of the family Rickettsiaceae (Proteobacteria), which includes only obligate intracellular symbionts and has often been reported in ciliates [25, 60,61,62], are abundant (> 5%, up to 42%) in some ciliates (e.g., cells 66 and 73, assigned to Paruroleptus and Trithigmostoma, respectively), but rare (< 0.1%) or undetectable in controls (Fig. 3). Bacteria of the genus Polynucleobacter (Proteobacteria), well studied as symbionts of freshwater ciliates, are completely absent from marine libraries, and their average relative abundance in freshwater controls is only 0.13 ± 0.11%, compared to up to 18.6% in freshwater ciliate microbiomes (cells 25, 76, and 79, assigned to the genus Euplotes; cells 12, 22, and 37, likely belonging to the genera Frontonia, Oxytricha, and Metopus, respectively). A taxon belonging to the family Neisseriaceae (Proteobacteria), which includes many commensal and parasitic bacteria, was only detected in the microbiome of ciliate 13 (likely affiliated to the genus Spirostomum), with a high relative abundance (20.13%). Bacteria of the genera Wautersiella (Bacteroidetes), Corynebacterium (Actinobacteria), and Ochrobactrum (Proteobacteria), all usually found in the microbiomes of humans and other mammals, were undetectable in controls but showed relative abundances up to 38.5%, 18.2%, and 22.2%, respectively, in several freshwater ciliate microbiomes (Fig. 3). Other prokaryotes were, on the contrary, present only in controls, like several taxa belonging to the order Bacteroidales (Bacteroidetes) (Fig. 3). Taxa retrieved with high relative abundances both in controls (> 1%) and in microbiomes (> 5%) included marine gammaproteobacteria of the genus Glaciecola (Proteobacteria) [63] (Fig. 3) and taxa of the order Sphingobacteriales (Bacteroidetes).

Fig. 3
figure 3

Different patterns of taxa relative abundances. Histograms showing three different patterns of abundances when comparing ciliate microbiomes (CM) and control communities (CC). Known symbiont: bacteria of the family Rickettsiaceae (Alphaproteobacteria), previously reported as symbionts of ciliates, show a high relative abundance (> 5%) in two ciliate microbiomes, while they are absent or negligible (< 0.1%) in control communities. Putative food source: bacteria of the genus Glaciecola (Gammaproteobacteria) are present in libraries from marine samples; their relative abundance is high both in marine ciliate microbiomes (> 5%) and in marine control communities (> 1%). This pattern is consistent with that of a food organism. Putative symbiont: bacteria of the genus Wautersiella (Flavobacteria) display the same pattern shown by known symbionts; therefore, they could be regarded as potential candidate symbionts of ciliates. Environmental bacteria: bacteria of the order Bacteroidales (Bacteroidetes) are present and abundant only in freshwater control communities. They are undetectable in ciliate microbiomes, indicating they are not associated with these protists. Y axes represents relative abundances, expressed as sequence percentages in the various libraries

Full size image

Genus-level taxa with the same abundance pattern as the known symbionts (>5 % in any microbiome library, < 0.1% in every control library) account for up to 66% of the total prokaryotic taxa in the microbiome of analyzed ciliates. This value varies broadly both for freshwater ciliates (0 to 66%) and marine ciliates (0 to 60%). When inspecting the number of putative symbiotic genus-level taxa within each prokaryotic phylum, most of the genera with this characteristic belong to the phylum Proteobacteria (21), followed by Actinobacteria (9), Firmicutes (8), and Bacteroidetes (4).

Discussion

The described approach provided reliable and satisfactory results on all ciliate specimens, regardless of their original environment and taxonomic affiliation. Rich and distinctive V3–V4 libraries were obtained with relatively little time investment, and microbiome profiles (lists of prokaryotes associated with various degrees with eukaryotic cells) could be unambiguously linked to molecularly identified hosts.

Although the sampling effort needs to be improved in order to obtain more reliable data, this preliminary work indicates that prokaryotic microbiomes of ciliate cells are different from total prokaryotic communities in the same site and habitat. As expected, background environmental communities have higher biodiversity, measured either as taxa richness or taxa evenness, which is also influenced by the presence of a community of active protistan grazers [64,65,66,67]. PCoA analyses and Permanova tests show that microbiomes of ciliates from the same habitat form a well-defined cluster, suggesting similar selective pressures on ciliate-associated prokaryote communities. These data suggest that, as already assessed for many macro-organisms like plants or metazoans [3, 4, 30, 31], ciliated protists possess a specific microbiome, distinguishable from the microbial community of the surrounding environment.

In our survey, differences between ciliate microbiomes and environmental communities were far less pronounced in marine than in freshwater samples. We found instead consistent differences between microbiomes of freshwater and marine ciliates, with habitat clearly surpassing other factors in affecting microbiome composition. However, these trends need to be tested with specifically designed studies and with a larger sampling effort, especially for the marine habitat.

Inspecting relative abundances of bacterial taxa in ciliate microbiomes and controls allowed to easily spot previously known bacterial symbionts of ciliates. All of them were present with relative abundances lower than 0.1% in control communities and higher than 5% in the microbiomes of their known hosts. Only bacteria of the genus Polynucleobacter appeared with slightly higher relative abundance values also in control communities (between 0.11 and 0.26%), and this can be explained by the fact that the genus comprises free-living as well as symbiotic strains [68]. Therefore, bacterial taxa showing similar abundance patterns can be reasonably regarded as putative symbionts, although more studies are certainly needed to refine these rough estimates. We are aware of no comparable data from single ciliate cells, but a recent meta-analysis screening of available libraries for a bona fide ciliate symbiont sequence also reported values well below 0.1% in reads from environmental libraries [69].

Using this criterion, known prokaryotic symbionts have in fact been detected here for the first time in association with unexpected hosts. For example, Polynucleobacter bacteria have been since reported as symbionts only in the genus Euplotes [22, 70], but were retrieved with high abundances (> 5%) also in the microbiomes of ciliates affiliated to the genera Frontonia, Oxytricha, and Metopus. Similarly, obligate intracellular bacteria of the family Rickettsiaceae were previously characterized in the ciliate genera Euplotes, Paramecium, Spirostomum, Diophrys, and Pseudomicrothorax [25, 60, 61]. Here, they have been additionally detected in microbiomes of ciliates affiliated to the genus Paruroleptus, which was never screened before for the presence of symbionts, and to the genus Trithigmostoma, in which macronuclear, unidentified symbiotic bacteria have been reported once [27]. It has to be mentioned that methanogenic archaea, well known as symbionts of anaerobic and microaerofilic ciliates [71], have been retrieved only in the microbiome of cell 37, assigned to the genus Metopus (one of the known hosts), and only in low relative abundances (one genus-level taxon at 1.94% and a second at 4.14%, both belonging to the Euryarchaeota lineage). They were not retrieved in other specimens belonging to putative hosts (i.e., a second Metopus and a specimen of Caenomorpha). The choice of amplification primers, targeting only a small fraction of archaea (426/160,767 according to Ribosomal Database Project), could have played a crucial role [72]. On the other hand, a previous investigation on the prokaryotes associated with a new species of the genus Metopus also failed in detecting methanogens [43], even if performed using archaeal-specific primers.

In addition to previously known symbionts, several novel prokaryotic taxa potentially associated with ciliates were here detected for the first time. Interesting examples include a taxon belonging to the family Neisseriaceae and bacteria of the genera Wautersiella and Corynebacterium. Protists in general, and ciliates in particular, have been already shown to harbor symbionts belonging to groups that also include pathogenic bacteria [61, 73,74,75,76]; the detection of a bacterial family including important mammal pathogens [77] and of genera mostly collected from human patients [78, 79] is definitely worth further investigations.

Obviously, prokaryotic taxa that are not strictly associated with ciliates are also bound to appear in single-cell libraries. Most commonly, they will be contaminants from the surrounding environment, or ingested organisms in food vacuoles. In either case, the involved taxa should also display high abundances in the total environmental community, except for ciliates with more selective feeding behaviors that might selectively prey on specific prokaryotes. Although food selection in filter-feeding ciliates is not entirely understood [80,81,82,83], the most important factor is generally considered to be prey size [84, 85]. Hence, in most situations, a comparative assessment of relative abundances will highlight the most promising symbiont candidates. Within our data collection, on the basis of results obtained for known symbionts, taxa with a relative abundance higher than 5% in host-associated libraries and lower than 0.1% in controls may be reasonably considered putative symbionts. Anything represented by 1% or more of control sequences should be treated carefully, even when present in high abundance in one or more ciliates. Borderline cases can certainly occur and should be tackled on a case-by-case scenario. Additionally, despite the depth of screening allowed by high-throughput sequencing techniques, the possibility that the used approach was not always sufficient to detect all bacteria associated with any single ciliate cell cannot be completely ruled out.

As might be expected, most of the prokaryotic genera with abundance profiles suggesting a closer relationship with the ciliate belong to phyla (Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes) that are the most represented both in the investigated environment and in the literature on symbionts of ciliates. In each ciliate-associated microbiome, the percentage of genus-like taxa that might represent true symbionts based on their abundance is very variable. This fits the expectation that many, but by no means all, ciliates harbor symbiotic bacteria [27]. It can be expected that, when one or few populations of stable symbionts are actually present, they will constitute a major part of the single-cell library. Conversely, in their absence, libraries will include a more diverse assemblage of prokaryotes with low abundances, only loosely associated with the eukaryotic cell.

Overall, the results of this pilot study provide an optimistic picture of the feasibility of single-cell microbiomics for ciliates. The method detected known symbiotic taxa in previously characterized as well as novel hosts and provided evidence for putative new symbionts that deserve attention. Studies applying the methodology to larger sample numbers should be able to address questions that are outside the possibilities of culture-dependent methods. To begin with, data on the environmental distribution of symbioses in ciliate populations could be collected, whereas information was previously confined to the few ciliates from each environment that could be cultivated. The temporal dynamics of symbionts in natural populations could be monitored. Once enough data is collected, correlations with abiotic parameters might also become apparent. As a consequence, the varying effect of symbionts on host fitness depending on the environment, currently only hypothesized based on lab observations [33, 35, 36], could be tested by properly designed studies in the field.