OUP user menu

Sample pooling obscures diversity patterns in intertidal ciliate community composition and structure

Matthias Engel, Anke Behnke, Sara Bauerfeld, Christine Bauer, Christian Buschbaum, Nils Volkenborn, Thorsten Stoeck
DOI: http://dx.doi.org/10.1111/j.1574-6941.2011.01255.x 741-750 First published online: 1 March 2012


The aim of this study was to assess the effect of sample pooling on the portrayal of ciliate community structure and composition in intertidal sediment samples. Molecular ciliate community profiles were obtained from nine biological replicates distributed in three discrete sampling plots and from samples that were pooled prior to RNA extraction using terminal restriction fragment polymorphism (T-RFLP) analyses of SSU rRNA. Comparing the individual replicates of one sampling plot with each other, we found a differential variability among the individual biological replicates. T-RFLP profiles of pooled samples displayed a significantly different community composition compared with the cumulative individual biological replicate samples. We conclude that sample pooling obscures diversity patterns in ciliate and possibly also other microbial eukaryote studies. However, differences between pooled samples and replicates were less pronounced when community structure was analyzed. We found that the most abundant T-RFLP peaks were generally shared between biological replicates and pooled samples. Assuming that the most abundant taxa in an ecosystem under study are also the ones driving ecosystem processes, sample pooling may still be effective for the analyses of ecological key players.

  • ciliates
  • community patchiness
  • intertidal sediment
  • microbial ecology
  • pooling
  • sampling strategy
  • T-RFLP


Sediment and soil are among the most complex microbial habitats on Earth (Urakawa et al., 1999). Their special structure represents a mosaic of microenvironments differing in physical, chemical, and biological properties, and consequently provides many different niches in which microorganisms are heterogeneously distributed at microscale level (Ranjard & Richaume, 2001; Franklin & Mills, 2003). Sediments are considered as sites for the accumulation and mineralization of organic matter (Heip et al., 1995; Middelburg et al., 1996), and unicellular eukaryotes (protists) are an important group of soil- and sediment-inhabiting organisms. They are essential components of the sedimentary microbial loop (Clarholm, 1994), as they play crucial roles in energy flow and elemental cycling (Jousset et al., 2010). For example, the degree of nutrient recycling is not only influenced by external factors like climate and input of plant and animal residues, but also internally by the community of protozoa in addition to prokaryotes, fungi, and metazoans, living in pores between water-covered soil aggregates (Bamforth, 1995). Moreover, protists like flagellates, naked amoebae, and ciliates are important bacterial grazers and they can enter tiny spaces unavailable to metazoans, like nematodes, feeding on bacteria (Bamforth, 1995). Still, there is only little knowledge about how these microbial eukaryotes are organized into functional assemblages (Caron et al., 2009) and even less about their spatial distribution and factors shaping distribution patterns (Finlay et al., 2000; Schwarz & Frenzel, 2003).

The pronounced small-scale variability in soils and sediments complicates investigations of microbial communities as it requires an adequate sampling strategy to obtain a representative sample that allows for a meaningful picture of community membership and structure. A common sampling strategy is to gather many smaller samples from various locations and pool them into a single, putatively representative, homogenous sample, which is then subsampled for further analyses. Previous studies investigated the effect of sample sizes on community profiles of bacteria and/or fungi and suggested a minimum of 1–2.5 g of soil for a reliable study (Ellingsøe & Johnsen, 2002; Ranjard et al., 2003; Kang & Mills, 2006). However, for a complete inventory of microbial diversity, the investigation of several smaller samples was recommended (Ranjard et al., 2003). Additionally, in molecular diversity studies, the effect of two different pooling strategies, prior to DNA extraction vs. prior to PCR, on microbial profiles derived from automated ribosomal intergenic spacer analysis (ARISA) was examined by Manter et al. (2010). It turned out that at any step pooling reduced the detected total phylotype richness, but the effect was different for the investigated sites and also for the taxonomic groups taken into account, namely bacteria and fungi (Manter et al., 2010). Such data demonstrate the necessity to develop and test sampling strategies for individual taxon groups. To the best of our knowledge, there are currently no molecular studies available testing the effect of sample pooling on microbial eukaryotes other than fungi. However, based on morphological inspection of marine plankton samples, Dolan & Stoeck (2011) demonstrated a differential variability in measures of species richness and community composition of tintinnid ciliates through repeated sampling, indicating the importance of an adequate sampling strategy in microbial ecology and diversity research.

The goal of this study was to assess the effect of sample pooling on phylotype richness and community structure of benthic intertidal ciliates in order to propose a sampling strategy for future studies in this field of research. Therefore, we profiled ciliate communities hailing from an intertidal sediment in the German Wadden Sea (Königshafen, Sylt, Germany) using terminal restriction fragment length polymorphism (T-RFLP) analyses of the SSU rRNA gene.

Materials and methods

Sampling site and procedure

All samples were taken on July 29, 2009, from an intertidal sandflat in Königshafen, a tidal embayment at the northern end of the island of Sylt in the German Wadden Sea (55°02′N; 8°26′E; Volkenborn & Reise, 2007). The sampling site was in the mid-intertidal zone and is characterized by an emersion period of 6–7 h and a grain size median of 340 μm. Samples for this study were derived from three different sites of a two-factorial nested block design, and each of the three plots was 400 m2 (20 m × 20 m) in area (Fig. ). Sites were located at a transect parallel to the coastline, about 100 m apart from each other. From each site (1–3) three sediment samples (X.1–3, 1.8 mL) were taken randomly, resulting in nine biological replicates in total, each of which was stabilized and preserved with 6.8 mL of RNA stabilization reagent (RNAlater; Qiagen, Hildesheim, Germany).

Experimental setup for sampling at the intertidal sandflat at Sylt (German Wadden Sea, simplified after Volkenborn and Reise, 2007). Each sampling plot (circles) was sampled at three different locations (squares).

RNA extraction, pooling and transcription

Total RNA was extracted from sediment samples (1 mg) using Qiagen's AllPrep DNA/RNA Mini kit. We worked with RNA to target active (living) organisms and to avoid (or at least minimize) the amplification from allochthonous material like dead organisms or cysts (Stoeck et al., 2007; Alexander et al., 2009). To compare individual replicates with pooled samples, our strategy had the following design. For the first data set (=individual biological replicates), we extracted total RNA from all nine individual sediment samples (squares in Fig. ). For the second sample set (pooled samples), we prior to RNA extraction pooled the three replicate sediment samples (squares in Fig. ) of each of the three sampling plots (circles in Fig. ), resulting in the pooled samples: 1-pooled, 2-pooled, and 3-pooled. The integrity of the extracted RNA was checked with the RNA 6000 Pico Assay (Agilent Technologies, Waldbronn, Germany) and screened for DNA contaminations by PCR amplification using an eukaryotic-specific primer set EukA (Medlin et al., 1988) and Euk516R (Amann et al., 1990). The PCR amplification protocol consists of an initial denaturation (5 min at 95 °C) followed by 30 identical amplification cycles (denaturation at 94 °C for 30 s, annealing at 56 °C for 30 s and extension at 72 °C for 45 s) and a final extension at 72 °C for 5 min. Finally, the RNA that tested negative in the DNA-contaminated PCR was transcribed into cDNA with the QuantiTect Reverse Transcription kit (Qiagen) following the instructions of the manufactures, including the utilization of a primer mix comprising an optimized mixture of oligo-dT and random primers.

Oligonucleotide primers and PCR amplification

To amplify the SSU rRNA genes of the microbial community, we applied a PCR primer set specific for the phylum Ciliophora, consisting of the forward primer Cil-F and a combination of three reverse primers (Cil-R1-3) (Lara et al., 2007). Ciliates were chosen because they are a significant component of marine interstitial protist communities (Carey, 1992) and because most, if not all, hitherto sequenced ciliate species can be targeted with one specific PCR primer set. For T-RFLP analyses, the forward primer was modified at the 5′-end with the phosphoramidite dye 6-carboxyfluorescein. The PCR mixtures contained 10–20 ng of template DNA, 5 U of HotStar Taq DNA polymerase (Qiagen), 1× CoralLoad PCR buffer (containing 1.5 mM MgCl2), 200 μM concentrations of each deoxynucleotide triphosphate, and 0.5 μM concentrations of each oligonucleotide primer. The final volume was adjusted to 50 μL with sterile water. The ciliate-specific PCR protocol for SSU rRNA gene amplification consisted of an initial denaturation (5 min at 95 °C) followed by 30 identical amplification cycles (denaturation at 94 °C for 30 s, annealing at 58 °C for 45 s and extension at 72 °C for 1 min) and a final extension at 72 °C for 5 min. Finally, PCR results were checked by agarose gel electrophoresis (1%), and after purification (MinElute PCR Purification kit; Qiagen), the DNA yield was quantified photometrically with the NanoDrop 2000 Spectrophotometer of NanodropTechnologies (Wilmington, DE).

T-RFLP analyses

To minimize the risk of the formation of artificial terminal restriction fragments (pseudo-T-RFs) because of inefficient restriction digestion of single-stranded PCR amplicons, we digested PCR products with a single-strand-specific mung bean nuclease (Egert & Friedrich, 2003). PCR products were incubated with 1× mung bean nuclease digestion reaction buffer and 2 μL of the enzyme (10 units μL−1; New England Biolabs, Ipswich, MA) at 30 °C for 40 min. Addition of SDS (0.01% final) terminated the reaction. Purification of the PCR products was carried out by isopropanol precipitation. The DNA yield was quantified photometrically with the NanoDrop 2000 Spectrophotometer of NanodropTechnologies.

An important step for T-RFLP analyses is the choice of restriction enzymes. As the goal of our analyses was the detection of diversity and differences in community structure, we chose enzymes that produce a large number of T-RFs with a reasonable length distribution. For the selection of adequate enzymes, we analyzed 610 publicly available sequences from described ciliate species using T-RF-CUT (Ricke et al., 2005) implemented in the ARB software package (Ludwig et al., 2004). After in silico digestion of sequences, we chose three out of 20 tested enzymes (HaeIII, MboI, and RsaI). Restriction digestion was carried out in 10-μL reactions according to the manufacturers’ instructions and using 5 μL of the purified PCR product. Digestion products were desalted by isopropanol precipitation and resuspended in 10 μL of deionized sterile water. Each restriction digestion for each enzyme and sample was performed as a triplicate (10, 15, and 20 ng), each triplicate per enzyme hailing from a separate PCR reaction. Prior to the analysis of T-RFLP reactions using a 3730 DNA analyzer (Applied Biosystems, Carlsbad, CA) at Seq-It Laboratories (Seq-It GmbH, Kaiserslautern, Germany), purified digestion products were mixed with 7.5 μL of a HiDi formamide solution (Applied Biosystems) and 0.5 μL of a ROX-labeled MapMarker1000 (Eurogentec, Cologne, Germany).

Electropherograms derived from the T-RFLP runs were analyzed using GelQuest© (version 2.1.2.SequentiX-digital DNA processing; Klein Raden, Germany) using default settings except for the following parameters: smoothing width, 10; baselining width, 50; minimum peak height for T-RFs, 75, minimum peak height for marker, 200; and hyperbin width, 1.1. Finally, only those T-RFs were taken into account that contributed at least 1% to the relative fluorescence (rA, based on height of T-RFs) of a sample (Noll et al., 2005). The relative abundance (rA) of each T-RF was calculated as rA = ni × 100/N in which ni represents the peak height of one distinct T-RF and N is the sum of all peak heights in a given T-RFLP profile. rA values were determined for all T-RFs detected in a size range between 51 and 693 bp for a given T-RFLP profile. Finally, data from triplicate reactions per sample (Table S1) were assembled by calculating the averaged rA of each individual T-RF to generate consensus T-RFLP profiles.

Analyses of community structure and membership

To investigate the variability among biological replicates, we calculated and compared microbial incidences and abundances of each sample based on consensus T-RFLP profiles. Therefore, T-RFLP results from the different enzymes were combined and analyzed by means of principal component analyses (PCA) as implemented in the Canonical Correspondence Analysis software (canoco©; Biometris). Furthermore, abundance-based Jaccard similarity indices (unadjusted) were calculated with spade (Chao & Shen, 2003–2005), and a upgma bootstrap tree was performed using the paup* software package 4.0b10 (Swoffort, 2001).

To test for the effect of pooling, we analyzed and compared community membership and structure of individual biological replicates and pooled samples. Initially, we determined the number and proportion of T-RFs that are (i) exclusive to replicate samples, (ii) shared by replicate samples and pooled samples, and (iii) exclusive to pooled samples. Finally, statistical analyses should reveal whether biological replicates (squares in Fig. ) from one sampling plot (circles in Fig. ) were more similar to each other and the respective pooled sample than to biological replicates and pooled samples from other sampling plots. These analyses employed the Student's t-test with the Jaccard similarity matrix as input data.


Phylotype diversity as revealed by T-RFLP

In total, 34–36 phylotypes were detected in the analyzed sediment samples depending on the applied restriction enzyme (HaeIII, 36; MboI, 36; RsaI, 34) (Fig. ). The amount of obtained T-RFs per sample differed notably among the nine biological replicates as well as the three pooled samples. HaeIII generated 10–18 phylotypes for the 12 investigated samples, and variations considering MboI were higher, ranging between 7 and 18 detected T-RFs. The range of detected phylotypes even increased by investigating samples with RsaI, revealing only four T-RFs in sample 1.3 and 18 T-RFs in case of 1.2 for the ciliate community. The latter example demonstrates that there was no obvious trend that phylotype richness was correlated with the plots samples were taken from, as the largest difference for phylotype richness was observed between biological replicates from the same plot (plot 1, samples 1.2 and 1.3). This was not only true for the restriction digestion with RsaI but also for the total number of T-RFs (1.2, 54 T-RFs; 1.3, 21 T-RFs). Regarding pooled samples, the variation between samples was less pronounced, and total numbers of T-RFs ranged between 30 (1-pooled) and 37 (3-pooled).

Phylotype diversity as detected by T-RFLP analyses. Numbers of obtained T-RFs are given in total as well as for each of the nine biological replicates and for the samples pooled per plot. Colors refer to the three different restriction enzymes applied (HaeIII, black; MboI, white; RsaI, gray).

Inter-replicate analogy

A PCA was performed to investigate the relationships between the ciliate communities in the nine individual biological replicates based on the detected T-RFs. This PCA revealed pronounced differences in the ciliate community compositions (presence–absence and abundance of a T-RF considered, Fig. ) because all samples are more or less randomly distributed along the two gradients. For example, two replicates of sample plot 1 (1.1 and 1.3) are close to opposite extremes of the PCA 1 gradient, and two replicates of sample plot 3 (3.1 and 3.3) are very distant to each other along the PCA 2 gradient. The eigenvalue of PCA axis 1 (x-axis) runs to 0.572, and of PCA axis 2 (y-axis) to 0.098. In sum, both axes explain 67% of the variation. This PCA analysis largely excludes a geographic distance effect on the ciliate community composition in the nine samples under study.

Analysis of ciliate community composition by means of the principal component analysis of the nine biological replicates. Filled symbols refer to biological replicates from the first plot (1.1, 1.2, 1.3), empty symbols to biological replicates from the second plot (2.1, 2.2, 2.3), and gray symbols to biological replicates from the third plot (3.1, 3.2, 3.3). Different symbols represent different biological replicates: square, first biological replicate (X.1); triangle, second biological replicate (X.2), and circle, third biological replicate (X.3).

A upgma dendrogram (Fig. ), based on the Jaccardabundance index, supports this first notion obtained by the PCA (Fig. ): in no case, the biological replicates from the same sampling plot form a discrete cluster. This analysis demonstrates that biological replicates from the same plot are not more similar to each other than to biological replicates from another sampling plot. The spatial smaller-scale (tens of meters) heterogeneity in the ciliate community structure is as pronounced as on a larger scale (hundreds of meters, see Fig. ).

Analysis of ciliate community structures of the nine biological replicates displayed by a upgma tree. Numbers at tree branches refer to calculated distances between biological replicates (Marczewski–Steinhaus-distances), and numbers at tree leaves refer to the nine biological replicates.

Effect of replicate pooling on microbial diversity detection

The effect of pooling on phylotype richness was determined by comparing numbers of T-RFs shared by the respective three biological replicates within a plot and the corresponding pooled sample to those exclusive to either the individual biological replicates or the pooled samples. An average of only 40% of T-RFs was shared between biological replicates and pooled samples, that is, only 40% of T-RFs were detected in at least one replicate of a plot and the corresponding pooled sample of this same plot (Fig. ). When analyzing the biological replicates individually, the number of detected T-RFs was nearly twice as high as in the pooled sample. In contrast, only 8% of the observed T-RFs occurred exclusively in the pooled samples and escaped detection in individual biological replicates.

Distribution of T-RFs detected within the three plots. Community membership was analyzed based on the incidence of phylotypes (i) in only one of three biological replicates (green, yellow, and blue), (ii) in at least two biological replicates but not in the pooled sample (red), (iii) in at least one biological replicate and the pooled sample (white), and (iv) in the pooled sample only (black).

In addition to community membership, we also analyzed the structure of ciliate communities, benefitting from the relative fluorescence of T-RFs as a measure of relative abundances (rA, see section). Therefore, we determined the proportion of the signal shared by biological replicates and pooled samples (Fig. ). In all experiments, the relative abundance abets a shift toward shared phylotypes, and the representation of biological replicates by a pooled sample improved when focusing on relative abundances. The proportion of the relative fluorescence of shared T-RFs comprised up to 85% of the total signal. As the representation of individual samples by pooled samples is improved when considering the relative fluorescence of T-RFs, it stands to reason that abundant T-RFs are preferentially detected by pooled samples, whereas rare phylotypes are rather missed and below detection limit. Indeed, statistical analyses showed that as a rule, T-RFs that were detected in biological replicates but not in the respective pooled sample have a significantly lower relative fluorescence (average across the relevant replicates) compared with T-RFs detected simultaneously in replicates and the pooled sample (P ≤ 0.05). As a rule, < 10% of the T-RFs that were missed by the pooled samples were above the average rA of the T-RFs detected simultaneously in biological replicates and in the pooled sample.

Abundance distribution of T-RFs detected within the three investigated plots. Community structure was analyzed based on relative abundance of phylotypes present (i) in only one of three biological replicates (green, yellow, and blue), (ii) in at least two biological replicates but not in the pooled sample (red), (iii) in at least one biological replicate and the pooled sample (white), and (iv) in the pooled sample only (black).

To determine whether pooled samples of a specific sampling plot were more similar to their respective biological replicates rather than to biological replicates from other sites, we statistically analyzed abundance-based Jaccard indices calculated for all pairwise comparisons. We found that only one of the pooled samples (3-pooled) was significantly similar to the corresponding biological replicates (3.1, 3.2, 3.3, P ≤ 0.05). Interestingly, the overall similarity between the three different investigated plots is relatively high, when only the pooled samples of each plot are compared with each other. While the mean Jaccard index for inter-plot comparisons was only 0.38 on average when comparing the biological replicates of these plots to each other, similarity was significantly higher (0.61, P ≤ 0.05) when calculated with three pooled samples (1-pooled, 2-pooled, 3-pooled). As a result, biological replicates analyses suggested a higher heterogeneity in ciliate community structures of the three different sampling plots, while pooled samples obscured spatial patterns.


Our results demonstrate a high patchiness of ciliate communities in samples derived from intertidal sediment in the German Wadden Sea, an extreme environment owing to tidal influence, high salinity, solar radiation, and desiccation (Volkenborn et al., 2007). This finding was not unexpected, as sediments and soils in general are regarded as one of the most complex habitats (Urakawa et al., 1999) with a mosaic of microenvironments providing ground for a large variety of different communities (Franklin & Mills, 2003). Sandy habitats, like the intertidal sediments under study, can harbor a large diversity of microorganisms in close proximity as a function of physical characteristics (Kang & Mills, 2006). From ultrastructural studies of soils, it is known that especially ‘larger’ microorganisms, like fungi, ciliates and amoeba, exhibit a pronounced heterogeneous distribution (Foster, 1988). Owing to their size, they are confined to larger voids (Foster, 1988) leading to an uneven distribution of diversity and biomass (Horton & Bruns, 2001). However, such high variability is not necessarily a function of distance. Samples of closer proximity are not necessarily more similar to each other compared with those of a longer distance. This was also the case for the sediment under study, as for the investigated samples we did not observe a distance effect on ciliate community composition and structure (Figs and ).

Of course, this result could be an artifact of the relatively low resolving power of the T-RFLP fingerprinting technique. Rather traditional techniques like cloning and sequencing of complete SSU rDNA sequences of taxa present in a sample would identify base-per-base differences between taxa and thus provide higher resolutions. However, such strategies still are a laborious task, which hampers the general applicability to ecological studies and limits the number of samples that could be investigated (Countway et al., 2005). We here cannot exclude that sequencing would paint a different or more detailed picture of distance–decay relationships, but the T-RFLP method has been widely accomplished and proven its suitability for community comparisons (Osborn et al., 2000; Lueders & Friedrich, 2003; Countway et al., 2005; Danovaro et al., 2006, 2006; Avis et al., 2010). Thus, the lower resolution does not lessen the applicability for the rapid screening of communities (Euringer & Lueders, 2008). A shortcoming common to all PCR-based surveys including T-RFLP and clone library sequencing is putative biases like the formation of chimeras during PCR (Wang & Wang, 1996), and several studies suggest that such PCR biases can adulterate the true phylotype richness (Hugenholtz et al., 1998; Qiu et al., 2001; Acinas et al., 2005; Avis et al., 2010). However, along with the consensus of many molecular studies (Wang & Wang, 1996, 1997; Danovaro et al., 2006; Huber et al., 2009), in this study precautions like short amplicon size (Wang & Wang, 1996; Huber et al., 2009), high elongation time (Avis et al., 2010), as few amplification cycles as possible (Polz & Cavanaugh, 1998; Acinas et al., 2005), the digestion of PCR products with mung bean nuclease prior to the restriction digestion (Egert & Friedrich, 2003), and the replication and construction of consensus of T-RFLP runs were taken to minimize the risk of obtaining PCR artifacts.

Therefore, the results obtained here are crucial for the design of experiments testing ciliate biogeography hypothesis like taxa–area and distance–decay relationships. The pronounced patchiness of ciliate communities very likely is due to the fact that the different biological replicates were samples of a number of diverse microniches within the sediment. Therefore, it a priori is not to expect that the complete ciliate community could be represented by single samples. Such a spatial patchiness was demonstrated only recently for tintinnid ciliates in Mediterranean water samples (Dolan & Stoeck, 2011). For our study, however, we benefited from this spatial heterogeneity as it allows testing the effect of sample pooling on our perception of protistan diversity patterns. This is because a high spatial heterogeneity in the taxon composition of protistan communities requires either the analyses of several samples in order to obtain a representative picture of the interstitial ciliate community or pooling these replicates into one single sample prior to analyses. However, our data suggest that a major proportion of ciliate diversity escapes detection (at least with a molecular fingerprinting technique) if biological replicates are pooled prior to analyses. Only taxa that appeared as highly abundant (i.e. taxa present with high copy numbers) in the individual replicate samples were also recovered in the pooled sample, while the less abundant ones were not detectable anymore.

This finding is consistent with previous studies on bacteria and fungi, suggesting that sample pooling adulterates true richness in these taxon groups (Ellingsøe & Johnsen, 2002; Ranjard et al., 2003; Kang & Mills, 2006; Avis et al., 2010; Manter et al., 2010). Such studies conclude that for the assessment of taxon richness, small and multiple samples are a better choice (Kang & Mills, 2006; Manter et al., 2010) because minor populations may be masked by dominant species in large and pooled samples (Ranjard et al., 2003). This is especially (but not exclusively) true for molecular approaches that tend to be saturated by the most dominant species (Grundmann & Debouzie, 2000). In case of the ciliate communities in the intertidal sediment under study, up to 70% of the T-RFs had escaped detection if only pooled samples were analyzed. Therefore, our results are in line with the work of Avis and colleagues (Avis et al., 2010) who found that pooling resulted in a distorted description of an artificially constructed fungal community. Moreover, even when considering larger zooplankton the same rule seems to apply, as it was found that especially rare taxa are not well represented in pooled samples (Ohman & Lavaniegos, 2002). This allows for the conclusion that in any case sample pooling is inappropriate for the investigation of diversity.

On the other hand, analyses of pooled samples may still be an acceptable option, when targeting protistan (ciliate) key players in ecological processes. It is reasonable to assume that the most abundant taxa in an ecosystem are at the same time the most important (active) ones in the ecological and biogeochemical processes shaping this system. Therefore, when targeting the most abundant taxa in a specific habitat with the aim to study their functions and roles in this habitat, it may be sufficient to analyze one large and homogenized sample instead of numerous individual biological replicates from this habitat. This is supported by our finding that up to 85% of the ciliate community detected in the cumulative biological replicates was also represented in the pooled samples when abundance (measured as relative fluorescence of T-RFs) was taken into account. However, one has to be aware that the copy number of SSU rRNA genes varies widely among protists (Countway et al., 2005; Zhu et al., 2005; Auinger et al., 2008), and it was demonstrated that among eukaryotes in general, this number apparently correlates with genome size (Prokopowich et al., 2003). As the rRNA gene copy number does not translate into numbers of cells, it is difficult to draw conclusion about possible key players in a given environment based on abundance data retrieved by, for example, T-RFLP or sequencing (Countway et al., 2005). Therefore, the assumption that pooling of samples at least allows for the investigation of ecologically important species does not necessarily apply if working with DNA. The utilization of RNA, on the other hand, offers some opportunities. It allows for targeting living organisms and avoids (or at least minimizes) the risk of amplification from allochthonous material like dead organisms or cysts (Stoeck et al., 2007; Alexander et al., 2009). As a general rule in prokaryotes as well as in eukaryotes, not only the RNA content (Wagner, 1994; Milner et al., 2001; Buckley & Szmant, 2004) but also the content of ribosomes (Hallberg & Bruns, 1976; Kief & Warner, 1981; Nomura, 1999) increases with growth rate and decreases in starving cells. Therefore, those taxa that are metabolically active (and therefore probably of some importance for the ecosystem under study) will still be detectable in pooled samples if using RNA extracts to generate SSU rRNA gene data.

As is the case with most specific PCR primer sets (Stoeck et al., 2006), the ciliate-targeting primer set we used is not insusceptible to bias. For example, individual taxon groups like Mesodinium pulex and Myrionecta rubra, which are known for their highly divergent SSU rRNA genes (Johnson et al., 2004) compared with other ciliates, escape the detection of the primer set used in this study. Being aware that M. pulex and M. rubra are part of the North Sea plankton community (Lei et al., 2010), it cannot be excluded that PCR-based fingerprinting techniques miss some dominant members of the communities under study.

A further consequence of the pooling effect was that differences in ciliate community structure between the three investigated sites decreased significantly when considering pooled samples only, probably due to the pooling-associated loss of sample-specific taxa (Manter et al., 2010). Thus, pooling minimizes the variability between samples and may provide an impression of the overall community structure (Ellingsøe & Johnsen, 2002; Manter et al., 2010) while the study of smaller samples results in a more accurate view (Ranjard et al., 2003). However, by pooling samples, locally abundant but heterogeneously distributed organisms may become rare because of an unintentional dilution effect, and this dilution effect may be enhanced by the competitive nature of PCR (Manter et al., 2010). Unfortunately, even if one is only interested in the overall community structure, pooling always leads to a loss of information about spatial variation among samples and hampers calculations of true experimental variability, which is inalienable for comparisons of communities, for example derived from different sites (Prosser, 2010; Dolan & Stoeck, 2011). Heterogeneity can be an important attribute of an ecosystem system and may represent valuable information that goes beyond ‘distracting noise’ (Ranjard et al., 2003). Therefore, we, based on our data, recommend the analyses of small sample biological replicates to study ciliate diversity in a defined sampling plot using molecular tools.

Authors’ contribution

S.B. and C.B. contributed equally to this study.

Supporting Information

Additional Supporting Information may be found in the online version of the article:

Table S1. Relative abundance of T-RFs of 12 investigated sediment samples derived from an intertidal sediment in the German Wadden Sea (Sylt).

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


We thank M. Müller-Frey for his support during sampling. We thank D. Hepperle from SequentiX (GelQuest) for help and support. This study was funded by the Deutsche Forschungsgemeinschaft, grant STO414/3-1 to T.S. and a grant from the University of Kaiserslautern to M.E. We thank two anonymous reviewers for helpful comments on the final version of this manuscript.


  • Editor: Riks Laanbroek


View Abstract