Introduction

The differences in the ecological requirements between species are the basis of the niche theory [1], implying that community composition variation would be dependent on environmental variation. Thus, under the niche theory, different habitats would favour different species to coexist [2]. Environmental gradients are indeed recognized as playing a pivotal role in structuring microbial communities in diverse habitats [3, 4] and across all spatial scales [5, 6]. The importance of the environmental conditions in shaping microbial communities was already envisaged by Baas Becking [7]: “Everything is everywhere, but, the environment selects” [8].

The rationale behind the “Everything is everywhere” principle is that due to their small size, high abundance and the ability to encyst or produce spores, microorganisms would have such high dispersal rates that they would not be restricted by geographical barriers [9, 10]. Indeed, microbial dispersal limitation seems to be much lower than that of macroorganisms [11, 12]. Nonetheless, a decrease in the compositional similarity with increasing geographic distance has also been found for microbial communities [13, 14]. These results suggest that biogeographic patterns, as found for macroorganisms, could also hold for microorganisms [15–17].

Rather than questioning whether microorganisms display patterns that resemble that of macroorganisms, now, we should focus on answering questions such as at what spatial extent the biogeographical patterns of microbes are comparable to those of macroorganisms? [16]. When considering multiple spatial scales, it is possible to investigate the relative importance of niche processes [18, 19], taking into account that dispersal limitation becomes more evident at large spatial extents [20]. Yet, not only spatial but also temporal scales should be taken into account [21], especially in ecosystems where environmental conditions are known to change seasonally [22]. In floodplains, periodic flood events can be considered as a disturbance, potentially changing the relative importance of deterministic and stochastic processes in community assembly [23], and are expected to homogenize the environments [24].

Another relevant issue is that different parts of a community appear to be assembled by different mechanisms [25]. Through a deconstructive approach, it is possible to disentangle the responses of groups of species with similar traits, such as dispersal ability, body size or habitat specialization [26–30]. For instance, dividing the community between common and rare species may better indicate the main processes accounting for species distribution patterns ([31, 32]; but see [33]). Chase et al. [34] suggested that the distribution of common and rare species is likely to be constrained by different assembly processes. They also suggested that, similarly to what is found in population genetics, rare species may be more prone to ecological drift (due to a smaller population size), while common species may be more subject to competition. In a metacommunity perspective, distributions of rare and common species are predicted to be more related to spatial and environmental processes, respectively [34].

Considering the species sorting perspective of the metacommunity theory [35], both biotic and abiotic variables are important in determining community compositions. Although most studies so far consider only abiotic variables in the environmental model, it has been shown that the inclusion of biotic variables considerably improves the explanatory power of the analyses [6, 36, 37]. Because of direct and indirect effects, biotic interactions should be considered in order to better understand the influence of dispersal on community assembly at various spatial extents [38] and the overall response to environmental gradients.

We gathered a large-scale dataset in four floodplains to investigate the spatial variation of ciliate metacommunities during different water level periods. We hypothesised that within floodplains, common species would be mainly related to environmental variables, due to their high abundance and dispersal rates, and rare species to spatial variables, due to smaller population sizes. We also expected a temporal variation on the contribution of the spatial and environmental components, such that during high water periods, we predicted a negligible importance of spatial variables within floodplains because, at this spatial extent, floods increase the connectivity between the floodplain habitats. When compared to the low water period, we also expected a lower contribution of environmental variables during the high water period, because the increase in connectivity would mask the effects of species-sorting processes (a pattern analogous to the one expected under the mass effect perspective). Alternatively, this result would derive from the homogenization effects of floods that reduce the range of environmental gradients [24]. During low water periods, lakes are more isolated and environmentally heterogeneous, and therefore, we expected an increase in the importance of both environmental and spatial variables. In addition, we expected a strong and preponderant biogeographic signal (as proxied by floodplain identities), regardless of the period and of the group analysed (common or rare taxa), since the spatial extent of our study should be large enough for dispersal limitation to occur.

Materials and Methods

Study Area

We sampled from 15 to 20 lakes in each of the following neotropical floodplains: Pantanal (18°46′–19°34′ S and 56°58′–57°46′ W), Paraná (22°40′–22°50′ S and 53°10′–53°30′ W), Amazon (3°02′–3°34′ S and 59°38′–60°50′ W) and Araguaia (12°49′–13°25′ S and 50°28′–50°43′ W), in both low and high water periods during 2011 and 2012. Those floodplains are widely distributed in the Brazilian territory, comprehending a wide latitudinal gradient (∼3000 km; Fig. 1).

Fig. 1
figure 1

Study area showing all sampling sites within each floodplain

Full size image

Sampling and Laboratory Analysis

Water samples for the analyses of the biotic variables, including all microbial communities (bacteria, heterotrophic flagellates and ciliates abundances) and chlorophyll-a concentrations, were taken at the subsurface (approximately 30 cm below the air-water interface) at the central, deepest region of each lake using polyethylene flasks. For zooplankton analyses, we filtered 600 L of water using a pump and a plankton net (68 μm). We analysed the ciliates in vivo using the subsampling technique [39], by passively concentrating 5 L of water samples into 100 mL using a plankton net (10 μm) and counting 10 replicates of 100 μL drops per site. Qualitative samples were also analysed using Sedwick-Rafter chambers to register rare species that could not be found during counting, until no new species were found. Ciliates were identified using optical microscopes (Olympus CX-41- brightfield and Olympus BX-51 - interference contrast system DIC) at a magnification of ×100–400, based mainly on the works of Foissner et al. [40], Foissner et al. [41], Foissner et al. [42], Foissner and Berger [43], Berger [44], Foissner et al. [45] and Foissner et al. [46]. Bacteria and heterotrophic flagellates were analysed from water samples treated with a fixative solution composed by alkaline Lugol’s solution, borate buffered formalin and sodium thiosulfate [47], filtered through black Nuclepore filters (0.2 and 0.8 μm, respectively) and stained with fluorochrome DAPI (4,6-diamidino-2-phenylindole; [48]). Quantification was done with an epifluorescence microscope at a magnification of ×1000 (Olympus BX51). Chlorophyll-a concentrations were determined in the laboratory [49]. Zooplankton total abundance was estimated, according to Bottrell [50], by counting at least three subsamples under an optical microscope at a magnification of ×40–400 depending on the taxonomic group (Olympus CX41).

Key environmental variables that are thought to structure ciliate communities were measured, including both abiotic (physical and chemical) and biotic (food resources and predators) variables. The following abiotic variables were measured in the field: dissolved oxygen (YSI portable oximeter), temperature (thermometer), water depth (portable depth sounder), pH, conductivity (Digimed portable potentiometers) and turbidity (LaMotte2008© portable turbidimeter). Water samples were measured for dissolved reactive phosphorus, nitrate and ammonia in the laboratory [51, 52]. For further analysis, all these dissolved nutrients were summed to create one variable (inorganic nutrients). We recorded the geographic coordinates of the sampling sites using a GPS.

Data Analysis

We partitioned the variance of three response matrices: whole community (all species), and the groups of common and rare ciliate species. We defined common and rare species using an adaptation of the quartile criterion [53, 54], in which the species are ranked from the most to the least abundant, but defining 25% of the most abundant as the common and 75% of the least abundant as the rare species, which seems to give similar information as other criteria (e.g. inflexion point; [33]). Using a partial redundancy analysis (pRDA; [55]), we tested the relative importance of three groups of explanatory variables (spatial, abiotic or biotic) in explaining ciliates community structure (for each period and group of species). The species matrices were Hellinger-transformed, allowing the use of Euclidean-based ordination methods (Legendre and Gallagher, 2001) [56]. We also tested for multicollinearity among explanatory variables, removing those with a variance inflation factor (VIF) value higher than 10 [57].

We applied the forward selection procedure [58] to the abiotic (dissolved oxygen, temperature, water depth, pH, conductivity, turbidity and inorganic nutrients) and biotic (bacteria, chlorophyll-a, heterotrophic flagellates and zooplankton) datasets to construct the models for each of those components. To evaluate the role of the spatial component, we built different models for the two spatial extents. The fine-scale spatial model (S) was based on the spatial variables produced by Moran’s Eigenvector Maps (MEM, [59]), derived from the geographical coordinates of the lakes (previously called principal coordinates of neighbour matrices (PCNM); [60]); because we wanted this component to reflect the within-floodplain variation, we constructed the spatial matrix based on MEM variables arranged in blocks, such that each block represented one floodplain and lakes from other floodplains received a value of 0. We used the R function proposed in Appendix 1 of the supplementary material in Declerck et al. [61] to calculate MEM variables. To construct the broad scale spatial component (F), we used a dummy variable representing floodplain identity (among-floodplains; [61]). We then evaluated the relative importance of the abiotic (A), biotic (B) and spatial components (S and F) in each hydrological period. The significance of each component was tested with 999 Monte Carlo permutations [62].

We further explored the relationship between common and rare taxa of ciliates and environmental variables, controlling for spatial relationships (within and among floodplains) through RDA triplots based on our results of variation partitioning. The same procedure was done to explore the relationship between common and rare taxa of ciliates and floodplain identities (largest spatial extent), controlling for environmental relationships and spatial relationships at the smallest spatial scale (within floodplains).

All the statistical procedures were performed using the libraries “vegan” [63] and “PCNM” [64] of software R 3.1.3 [65].

Results

Abiotic and Biotic Variables

Of the several abiotic variables measured, the most conspicuous variation between the hydrological periods was registered for depth: During the high water period, Amazonian lakes reached more than 15 m, with a mean value of 6 m considering all floodplains; on the other hand, during the low water period, mean values were around 2 m, with a maximum depth of 7.2 m. Some biotic variables also showed great variation. For example, much higher values of chlorophyll-a and zooplankton abundance (rotifers, cladocerans and copepods) were registered during the low water period (Table 1).

Table 1 Mean, minimum (min) and maximum (max) values of abiotic and biotic variables registered in lakes from all floodplains during two hydrological periods
Full size table

Common and Rare Ciliate Species

Out of the 22 species considered as common in the study, most taxa were shared between the two hydrological periods (Fig. 2). Oligotrichous ciliates were by far the most representative, with seven species being considered common. Among them, Halteria grandinella, Tintinnidium sp. and Tintinnopsis sp. showed particularly high abundances in both periods (Table S1). As for the rare species, exclusive taxa in each period slightly exceeded the shared part (Fig. 2).

Fig. 2
figure 2

Relationship between the density of common and rare taxa of ciliates during the high and low water periods

Full size image

Variation Partitioning

Most of our predictions were supported by the data (Table 2). First, common species were mainly related to abiotic (A) and biotic (B) components (environmental variables), but not significantly related to the spatial component within floodplains (S). For rare species, during the low water period, we detected a significant adjusted coefficient of determination (adj. R 2) associated to the spatial component S (within-floodplains) and non-significant coefficients associated to A or B. However, during the high water period, rare species were significantly related to environmental gradients (B), contrary to what we expected. Second, the prediction of higher adj. R 2 (sum of the pure components A, B and S) during the low water period than during the high water period was also supported. For rare species and for the whole community, our prediction of a high importance of S during the low water period, but a negligible importance during the high water period, was also supported. Finally, we found that the spatial component among floodplains (F) was the component with the highest adjusted coefficient of determination regardless of the response matrix and hydrological period (Table 2).

Table 2 Variation partitioning of ciliate community data in relation to pure abiotic (A), biotic (B) and spatial variables (S and F) during two hydrological periods
Full size table

Common and rare species were influenced by different environmental variables (Table 2). Abiotic variables were important to explain the variation of common species during both hydrological periods. Temperature and turbidity were selected for both periods, while depth was included in the abiotic model only during the high water period and both dissolved oxygen and conductivity were exclusive of the low water period. The biotic component was important for rare species during the high water period (copepods density) and for common species in the low water period (chlorophyll-a, bacteria and rotifers density). Total species matrix was always explained by the same variables as the common or rare species, when the same component was concomitantly significant.

RDA triplots based on the relationship between ciliates and environmental variables of variation partitioning (controlling for spatial relationships) showed that the variable most strongly associated with the common taxa was turbidity during the high water period (mainly related to Tintinnidium sp.) and conductivity during the low water period (mainly related to Tintinnidium sp., Urotricha farcta and Euplotes sp.; Fig. 3a, b). Moreover, biotic variables, such as rotifers and chlorophyll-a, were also important for common taxa during the low water period (Fig. 3b). For rare taxa, copepods were positively associated with Paramecium aurelia and Limnostrombidium sp. and negatively associated with Colpidium kleini and Urocentrum turbo (Fig. 3c).

Fig. 3
figure 3

RDA triplots based on the relationship between the common taxa during the high water period (a), common taxa during the low water period (b), and rare taxa during the high water period (c) and the environmental variables, based on results of variation partitioning. Ar Actinobolina radians, Bp Balanion planctonicum, Bsp Burselopsis sp., Cc Codonella cratera, Cg Cyclidium glaucoma, Ch Coleps hirtus, Ck Colpidium kleini, Cp Carchesium polypinum, Cs Colpoda steinii, Dn Didinium nasutum, Eg Enchelys gasterosteus, Esp Euplotes sp., Fa Frontonia atra, Fl Frontonia leucas, Hg Halteria grandinella, La Laginophrya acuminata, Ll Lembadion lucens, Lsp Limnostrombidium sp., Lv Limnostrombidium viride, Mp Mesodinium pulex, Pa Paramecium aurelia, Pe Paradileptus elephantinus, Psp Paradileptus sp., Rh Rimostrombidium humile, Rl Rimostrombidium lacustris, Sv Stokesia vernalis, Tp Tetrahymena pyriformis, Tisp Tintinnidium sp., Tsp Tintinnopsis sp., Uf Urotricha farcta, Ut Urocentrum turbo, Va Vorticella aquadulcis, Vn Vorticella natans, Vsp Vorticella sp.

Full size image

Site scores from partial RDA of the relationship between ciliates and floodplain identities (controlling for environmental and spatial relationships within-floodplains) further evidenced that, although variation in both common and rare species were more associated with floodplain identities (F; see Table 2), they showed different responses. Essentially, separation by floodplains was less evident for common species (Fig. 4a, b) than for rare species (Fig. 4c, d). Separation for rare taxa during the low water period (Fig. 4d) was the most noticeable, with Burselopsis sp., Phialina pupula and Uronema nigricans distinctly associated with Araguaia floodplain, whereas Rimostrombidium lacustris and Limnostrombidium sp. were associated with Paraná River floodplain.

Fig. 4
figure 4

RDA triplots based on the relationship between the common taxa during the high water period (a), common taxa during the low water period (b), rare taxa during the high water period (c), rare taxa during the low water period (d) and floodplain identities (spatial variables of the largest spatial extent: among-floodplains), based on results of variation partitioning. Bp Balanion planctonicum, Ch Coleps hirtus, Cp Carchesium polypinum, Dn Didinium nasutum, Fl Frontonia leucas, Hg Halteria grandinella, Lv Limnostrombidium viride, Mp Mesodinium pulex, Rh Rimostrombidium humile, Tp Tetrahymena pyriformis, Tisp Tintinnidium sp., Tsp Tintinnopsis sp.

Full size image

Discussion

Testing the Hypothesis That During the High Water Period, We Would Have Lower Adjusted Coefficients of Determination

The relative importance of the assembly processes of the planktonic ciliate communities was dependent on the hydrological period and on patterns of commonness and rarity. During the high water period, the spatial component (S: within floodplains) was not significant for any of the ciliate groups (common, rare and total); on the other hand, environmental components (either abiotic or biotic) were always significant. Therefore, we detected a predominant role of species sorting in structuring the ciliate communities during flood events. Those results contradict our expectations that an increase in hydrological connectivity, which enhances the rates of passive dispersal between floodplain habitats, would mask the effects of species sorting by allowing inferior competitors to persist in the community (mass effects). Similar results were found for zooplankton [66], bacteria [67] and phytoplankton [68] communities of the highly connected ponds of “De Maten” complex, where the environmental component always exceeded the spatial component. Those results could be explained by the very high population growth rates of small-sized organisms, which allows them to efficiently track environmental changes [67]. This environmental control would, then, be overwhelmed only at conditions of extremely high water flow events (i.e. large storm events: [69]), in which microbial turnover rates would not be sufficient to overcome the high rates of cell import.

Testing the Hypothesis That Rare Species Would Be More Related to Spatial Variables and Common Species Would Be Related to Environmental Variables

We found that most common species were present in both periods, whereas the majority of rare species were found in one particular hydrological period (Fig. 2). This indicates that few species were ubiquitous and temporally persistent [70]. In line with our prediction, those common species were related to environmental gradients [28, 71], during both hydrological periods, reinforcing the role of species sorting processes [6, 67]. Also, the spatial variables within floodplains (component S) were not significant. For rare species, we detected a sole and significant effect of spatial variables only during low water period, suggesting some degree of dispersal limitation within floodplains during low waters [30]. However, during the high water period, we detected a significant environmental effect for rare species. Biota exchanges among habitats caused by and during floods enhance the possibility of dispersal and colonization by rare species [24]. Possibly, the flood event mitigated the dispersal limitation that rare species of ciliates were undergoing in the absence of flood, which prevented species sorting to occur. This is consistent with the experimental results obtained by Declerck et al. [72], who showed that the strength of the relationship between microbial communities and environmental factors was increased with increasing dispersal intensity.

Floods also seem to affect the relative importance of environmental and spatial processes for common and rare taxa of benthic invertebrates in the Upper Parana River floodplain [29]. Temporal variations were likewise found to be important for driving dissimilar patterns of rare and abundant protists [70]. Thus, our results were partially in line with our expectations that rare species seem to be mainly related to spatial variables, because we found that temporal disturbances (i.e. flood) can alter this pattern. Nevertheless, in the absence of disturbance, this pattern seems to apply to rare ciliates in floodplain lakes, as well as for marine ciliates [32, 71], bacteria [73] and microbial eukaryotes [74].

Our results also indicated that the nature of the species sorting processes differed between common and rare species. For instance, we found that copepods density was the main predictor of the abundance of rare species during the high water period. If the relationship between rare ciliates and copepods emerges due to prey-predator relationships, at least for some species with negative scores (e.g. Colpidium kleini and Urocentrum turbo), then our results seem at odds with the view that predation is assumed to be important only for the most abundant taxa of ciliates [75]. According to this view, the selective nature of the predators results in a higher probability of finding the most abundant prey [76]. However, because copepods are composed of both filter-feeding (non-selective/generalists-calanoids) and raptorial (selective/specialists-cyclopoids) taxa [77], selective predation may not always occur. Thus, when predators are not selective, both common and rare species would be similarly affected [76]. On the other hand, the positive association between some species (e.g. Paramecium aurelia and Limnostrombidium sp.) and copepods density may indicate a shared response to environmental gradients [78]. Rotifer density was found to be the main variable accounting for the variation of the common species during the low water period. Predation, for ciliate species with negative scores (Tintinnidium sp. and Balanion planctonicum), and shared response to environmental gradients (e.g. chlorophyll-a; see Fig. 2b), for those with positive scores (Carchesium polypinum and Coleps hirtus), are also likely explanations for these results.

Abiotic factors, particularly temperature and turbidity, were also important in explaining the variation of common ciliate species, regardless of the hydrological period. Temperature is a key variable, as it is strongly linked to ciliate growth and feeding rates [79]. Although some species of protists have been found to be tolerant to experimental manipulations in turbidity [80], our results suggest that this variable cannot be discarded as important in structuring ciliate communities. Our results also highlight interesting uncertainties on how turbidity affects ciliates. For instance, during the low water period, Tintinnidium sp. density was highest in low turbidity lakes; however, during the high water period, this pattern was reversed. Conductivity was the second most important predictor of ciliate communities during the low water period. Küppers and Claps [81] also found a positive association between conductivity and two taxa shared with our study (Euplotes sp. and Frontonia leucas). They also found a negative association between Halteria grandinella and conductivity. However, while we found a positive association, Küppers and Claps [81] found a negative association between Coleps hirtus and conductivity. However, considering that both conductivity and turbidity are affected by other environmental processes (e.g. vertical mixing), these relationships do not indicate causality. In general, our results suggest that generalizations regarding the responses of ciliates to key variables (e.g. turbidity and conductivity) structuring plankton communities are still premature. In general, our results suggest that generalizations regarding the responses of ciliates to key variables (e.g. turbidity and conductivity) structuring plankton communities are still premature.

The divergent patterns found for common and rare species (common species more related to environmental variables and rare species to spatial variables) may be accounted for by differences in dispersal rates, which are mainly attributable to the abundance of the populations [9]. Thus, rare species, with smaller population sizes, would have lower dispersal rates than the common ones [82]. If we expect bad dispersers to be more affected by spatial distance [83–85], the same should hold true for rare species. Strong dispersal limitation is much less likely to occur for common species, and being stronger dispersers, they should be more subject to environmental control [86].

Testing the Hypothesis That There Would Be a Strong and Preponderant Biogeographic Signal

In addition to the significant contributions of the environmental variables, as described above, we found that floodplain identity was the strongest predictor of ciliate communities. The sole interpretation of the adjusted fractions (Table 2) would, therefore, indicate a preponderant biogeographical signal in our study. However, this is a necessary but not a sufficient condition to infer a biogeographic pattern (or “microbial provincialism” according to Martiny et al. [17]) and, at the same time, refute the Baas-Becking’s statement that “Everything is everywhere, but, the environment selects” (advocated by Fenchel and Finlay [10]). Aside from the significant adjusted R 2 associated to the variable floodplain identity, we should also observe a clear separation of the floodplains in the community space (see Fig. 1 in Martiny et al. [17]). At least for rare species, a combination of those results was found (see Fig. 3), indicating biogeographic patterns for rare taxa of ciliates in floodplain lakes. As discussed above, rare species are indeed expected to show dispersal limitation [82]. Several authors have suggested that protists display biogeographic patterns by finding significant distance-decay relationships (DDR) (e.g. [87, 88]); however, because DDR can also emerge due to environmental variation, those results are uncertain. In this way, our results for the ciliate community could clearly demonstrate microbial provincialism [17], together with others investigating pure spatial effects on protists (e.g. diatoms: [89, 90]). Further evidence of this is showed by the many flagship ciliate taxa [91] having restricted geographical distributions, which is advocated as being “an ultimate proof of endemism” [92], contradicting a cosmopolitan distribution.

Although microorganisms are known to exhibit very high dispersal rates due to small size and high abundance [9, 10], not all microbial species have the same probability to passively disperse. For example, airborne microbes of <20 μm were found to have a wider range of reach than the ones with <40 μm [93]. Considering that planktonic ciliates range from ∼15 to 200 μm in size [94], only a subset of the species would be easily dispersed at great distances. Moreover, after the microorganisms dispersal and arrival in the habitat, they need to encounter favourable conditions to reproduce and are subject to the competition exerted by the local communities. In this way, Weisse [95] called effective dispersal the “successful establishment of immigrants” as the result of the dispersal itself plus the abiotic and biotic constrains determining the observed distribution of species over space and time. This is in line with our findings, which suggests that even though ciliates may have a high potential dispersal (sensu [95]), that does not seem to be translated in an effective dispersal (sensu [95]), especially at larger spatial extents.

For common species, the separation of the floodplains in the community space was less conspicuous, suggesting a more ubiquitous distribution. Nevertheless, the fact that we could compare the spatial signal at two spatial scales, and it was only significant at the largest spatial scale, suggests a more limiting dispersal with increasing spatial extent [96]. Thus, the degree to which the spatial component was important depended on the spatial extent considered [20]. Common species are indeed expected to experience less dispersal limitation [82]. Thus, common species are closer to a cosmopolitan distribution, and this could partly explain the consensus among researchers: that microorganisms have wider geographic ranges than macroorganisms. An explanation may be due to the ancient ages of microbial species, which would have given them sufficient time to disperse [92].

In conclusion, we can view ciliate metacommunities as constituted by subcommunities of common and rare species which differ in their dispersal ability, ecological niches and, consequently, in their main assembly mechanisms. Our results partially agree with the perspective that common species are more close to be considered cosmopolitan, while rare species display more restricted distributions [73, 97, 98], because temporal and spatial scales may affect those patterns. All in all, microbial biogeography should not really differ from that of macroorganisms [15, 17].

Considering that we used morphological taxonomy in the identification and that some species remained undetermined (sp.), there are implications on the results that must be considered. Regarding the relationship between the community and environmental variables, the amount of variation explained by the environment should increase when closely related species are grouped together, since they usually share similar traits and have similar environmental preferences [99]. On the other hand, grouping of species may decrease the relative importance of spatial variables [100]. Fortunately, this effect is minimized across large geographic scales [101]. Given that our analysis detected a spatial signal, we believe that these issues likely did not affect the main patterns we found in our study. However, despite these inferences, we also have to emphasize the need for more studies to solve taxonomic uncertainties regarding ciliates.

It is interesting to note that the overall predictive power of our models was not very high (ranging from ∼6 to 30% of explained community variation). This is unlikely to reflect the lack of important contemporary variables sampled in our study, which incorporated several abiotic and biotic variables known to exert a strong influence on the ciliate community (see “Materials and Methods”). Instead, it seems to be an intrinsically characteristic of the small sized organisms which are usually less predictable than their larger counterparts [37, 102]. However, as the majority of studies so far, ours was performed with snapshot samplings, so we suggest that future studies take into account the past environmental conditions, which can better predict microbial spatial differences [103].