With substantial reductions in DNA sequencing costs combined with higher sequence yields, amplicon sequencing has revolutionized our view of microbial ecology (Sogin et al., 2006). It produces a culture-independent molecular characterization of the microbial community composition, and application of amplicon sequencing has successfully discovered novel microbes and characterized microbial diversity from a wide range of environments (Caron et al., 2012).

Due to its high specificity and sequence conservation, the 18S rRNA gene has become the most commonly used marker to explore eukaryotic protist community structure in both aquatic and terrestrial environments (Countway et al., 2005; de Vargas et al., 2015). However, 18S gene sequencing has inherent drawbacks that are typical of high throughput sequencing studies, such as PCR chimeras and sequencing errors (Caron et al., 2012). Ongoing research has focused on fixing such issues to provide a more accurate taxonomic description. Primers have been continuously modified and multiple hyper-variable regions can be simultaneously sequenced to reduce primer mismatch (Parada et al., 2016; Lin et al., 2017). Metagenomics has also been employed to preclude PCR-based bias (Eloe-Fadrosh et al., 2016).

Besides inherent technical issues, another source of bias to quantifying community composition with 18S gene sequences stems from variable copy numbers of ribosomal genes (Countway et al., 2005; Caron et al., 2012). Read counts of 18S genes are commonly used to estimate proportions of protists in amplicon sequencing analyses. However, the relative abundance of 18S gene copies in eukaryotic plankton collected from environmental samplescan be attributed both to variation in the relative abundance of different organisms, and to variation in genomic 18S copy number among those organisms (Zhu et al., 2005; Godhe et al., 2008). Phylogeny-based approaches have been developed to estimate ribosomal gene copy numbers for prokaryotes (Kembel et al., 2012; Angly et al., 2014), however, accuracy of such estimation can be compromised for protists due to the limited number of genomes that have been sequenced. Zhu et al. (2005) have developed a quantitative PCR-based approach to estimate 18S gene copy number for picoeukaryotes by normalizing total copies of 18S gene in the sample with cell abundance, but the results are highly dependent on DNA extraction efficiency, primer specificity and cell enumeration, and can be impractical for uncultured but prevalent phytoplankton species. Estimating 18S gene copy number remains an arduous task.

Due to the large number of sequences produced through high-throughput sequencing, a computational method that determines gene sequencing coverage has been recently developed and has become a promising method to estimate gene copy number variations (Zhao et al., 2013). It has been successfully applied to bacterial communities (Perisin et al., 2016), however, the application of such an approach on protist communities has yet to be performed. By normalizing sequencing coverage of 18S genes with that from single copy genes, we quantified 18S gene copy numbers of selected marine eukaryotic phytoplankton species whose draft assemblies are available in NCBI’s GenBank database. Thus far our results provide 18S gene copy number estimates for multiple representative species from four common phytoplankton classes and found large interspecies and strain-level 18S rRNA gene copy number variations across the different phytoplankton species.

Estimates of mean 18S V4-region gene copy numbers range from approximately 2–166 across the seven closed/draft phytoplankton genomes (Figure 1 and Supplementary Table 1). O. tauri, on the lower end, was estimated to have, on average, 3.4 copies of the 18S gene across 13 strains after GC% correction to account for possible sequencing bias. This is similar to the previously reported four copies of the ribosomal gene in strain RCC4221 (Blanc-Mathieu et al., 2014). GC% was shown to have a significant correlation with sequencing depth in O. tauri genome assemblies and could account for the observed difference from the reported four 18S gene copies estimated by Blanc-Mathieu et al. (2014) (Supplementary Figure 3). On the higher end, our results suggest an average of 160 18S gene copies in the dinoflagellate S. kawagutii, which could be attributed to large and repetitive genomes typical of dinoflagellates (Lin, 2011; Wisecaver and Hackett, 2011; Shoguchi et al., 2013; Lin et al., 2015). The notable two orders of magnitude difference can result in highly biased phytoplankton composition characterization, over/under-estimating species with higher/lower 18S gene copy number. For example, a simulated community with equal 18S gene sequence abundances from seven representative species seems to suggest that each species contributes equally to the community. However, upon 18S gene copy number correction, O. tauri and P. tricornutum actually dominate the natural community (Figure 2). In addition, strain level differences in 18S gene copy numbers were also detected, further highlighting the strong variation among eukaryotic phytoplankton, which may be a common characteristic among all protists. Estimates of 18S gene copy numbers in the coccolithophorid, E. huxleyi ranged from 16 to 109 across 14 different strains collected at different sites across the globe (Supplementary Table 1). Strains from English Channel have significantly higher 18S gene copies than those from the neighboring Bergen Sea (Supplementary Figure 4 and Supplementary Table 1). In contrast, 18S gene copy number was found to be more consistent for the diatom P. tricornutum strains that were also isolated from a large distribution of locations. It is unclear at this time whether this degree of variation in 18S gene copy number is due to inherent variability within particular phytoplankton groups or a result of the geographical location where the isolates were obtained (Figure 1). Further research is necessary to assess whether there are biogeographical patterns to these strain-level differences. The intraspecies geographic variation in 18S gene copy number adds another level of complexity to characterizing plankton community structure, and suggests that site-specific copy number estimations and corrections may be necessary for further compositional studies.

Tremendous sequencing effort has been devoted to estimate ribosomal copy numbers in bacteria and archaea, but our knowledge on their eukaryotic protist counterparts has benefited much less from the fast-developing sequencing technologies. Copy number variation can exert a strong influence on protist community composition and lead to biased ecological inferences. Recent bioinformatics technologies including metagenome assembly of genomes have the potential to provide new insights to estimating 18S gene copy numbers for multiple species simultaneously and will dramatically increase the number of estimates for protists (Delmont et al., 2018). With slight modifications, the bioinformatic pipeline implemented in this study can be applied to environmental sequences to provide an estimate of 18S gene copy numbers for dominant species. Assembled contigs that can be accurately taxonomically and functionally annotated as single copy genes, along with 18S genes, will generate an 18S gene copy number estimate. We propose that this pipeline be applied to metagenomic samples obtained from each location in which amplicon sequencing based compositional analyses are routinely performed.

Our findings emphasize the need to incorporate 18S gene copy number variation in protist compositional studies and provides a promising means to measure them in eukaryotic plankton, although further research is warranted due to complex genomes and polyploidy, especially in Alveolates. We anticipate that continuing sequencing efforts with consistent sequencing platforms and guaranteed sequencing depth along with emerging bioinformatics tools will add more perspectives to 18S gene copy number estimates and correction and will result in more accurate representation of eukaryotic community structure.

WG designed the research and analyzed the data. WG and AM wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.