DNA extraction bias is more pronounced for microbial eukaryotes than for prokaryotes

Abstract DNA extraction and preservation bias is a recurring topic in DNA sequencing‐based microbial ecology. The different methodologies can lead to distinct outcomes, which has been demonstrated especially in studies investigating prokaryotic community composition. Eukaryotic microbes are ubiquitous, diverse, and increasingly a subject of investigation in addition to bacteria and archaea. However, little is known about how the choice of DNA preservation and extraction methodology impacts perceived eukaryotic community composition. In this study, we compared the effect of two DNA preservation methods and six DNA extraction methods on the community profiles of both eukaryotes and prokaryotes in phototrophic biofilms on seagrass (Zostera marina) leaves from the Baltic Sea. We found that, whereas both DNA preservation and extraction method caused significant bias in perceived community composition for both eukaryotes and prokaryotes, extraction bias was more pronounced for eukaryotes than for prokaryotes. In particular, soft‐bodied and hard‐shelled eukaryotes like nematodes and diatoms, respectively, were differentially abundant depending on the extraction method. We conclude that careful consideration of DNA preservation and extraction methodology is crucial to achieving representative community profiles of eukaryotes in marine biofilms and likely all other habitats containing diverse eukaryotic microbial communities.


| INTRODUCTION
Advances in sequencing technology and paradigm shifts in microbial ecology have led to a prolific rise in studies that use metagenomic and marker gene polymerase chain reaction (PCR) amplicon sequencing to assess microbial communities in various environments. Essential to all of these efforts is the preservation and extraction of DNA from environmentally or host-associated microbial communities. It is well known that the choice of DNA preservation and extraction method can impact the perceived relative abundance of microbial taxa in microbial communities (e.g., Martin-Laurent et al., 2001). Differences in community composition depending on the DNA extraction method are referred to as extraction bias, which can have various causes, many of which are linked to the ability to lyse microbial cells (Koid et al., 2012). A wide variety of commercial kits and custom protocols have been MicrobiologyOpen. 2022;e1323. www.MicrobiologyOpen.com developed to provide representative and reproducible DNA extraction from different sample types. For some environments, extraction bias has been evaluated by comparing the outcome of different extraction protocols, in some cases, leading to general recommendations on method choice (e.g., Albertsen et al., 2015;Weber et al., 2017). A majority of existing studies have focused on prokaryotic communities, reflecting an emphasis on bacteria and archaea in molecular microbial ecology.
However, in most natural environments, microbial eukaryotes are abundant and diverse and play essential roles in ecosystem processes. Whereas they have traditionally been studied using microscopic methods, studies using molecular methods have revealed novel taxa that escape microscopic detection or identification (Jones et al., 2011;Liu et al., 2009). In the wake of numerous influential studies on prokaryote diversity in various ecosystems, microbial eukaryotes are receiving renewed attention by taking advantage of available high-throughput sequencing technologies (Delmont et al., 2022;Lima-Mendez et al., 2015).
Due to a high diversity of cell envelopes found in microbial eukaryotes, ranging from single membranes in ameboid protists to silica frustules of diatoms or thick cellulose cell walls of green algae, effective cell lysis and subsequent DNA recovery pose unique challenges. Despite this, extraction bias has so far received little attention in surveys of microbial eukaryotes (but see Donn et al., 2008;Koid et al., 2012;Mäki et al., 2017;Santos et al., 2015;Vesty et al., 2017). In addition, microbial eukaryotes and prokaryotes are intermingled in most microbial communities, and extraction methods that recover DNA well from a variety of eukaryotes and prokaryotes are needed to achieve an accurate representation of microbial community composition.
Here, we compared the effect of different popular commercial and custom DNA extraction methods on the perceived community composition of prokaryotes and eukaryotes in marine phototrophic biofilms growing on seagrass leaves. We aimed to assess whether extraction bias affects microbial eukaryotes and prokaryotes at a similar magnitude in the same environment and whether this bias depends on the sample preservation method.  Figure 1a). These two methods were characterized by more gentle lysis conditions, weak bead beating (smaller beads than in the other tested methods; see Table A1) and enzymatic lysis, compared to the other methods that use harsh bead beating, indicating that incomplete lysis of some eukaryotic cells may underlie the observed pattern. However, when investigating which eukaryotic taxa were differentially abundant in these methods, we found that metazoans,

| DNA yield does not impact community composition
The DNA yield differed significantly among extraction methods   Figure A5 for the remaining comparisons flash-frozen samples. DNA yield did not significantly explain the variation in perceived community composition across prokaryotic and eukaryotic samples (PERMANOVA, p > 0.2 and p > 0.05, respectively), indicating that factors that affect the overall yield are different from those giving rise to DNA extraction bias. This is reassuring since extraction yield can vary substantially even between replicate samples under the same extraction method (see e.g., PowerBiofilm method, Figure A1), but this does not compromise the reproducibility of community composition patterns (Vishnivetskaya et al., 2014).

| CONCLUSIONS
Most microbial DNA extraction methods have been developed and optimized for prokaryotes and may therefore be inadequate for microbial eukaryotes, which have a high diversity of cell envelopes, posing unique challenges for effective cell lysis and subsequent DNA recovery. It is unlikely that we will ever arrive at one optimal methodology that captures all organism groups without bias. It is also not the aim of this study to offer specific recommendations for DNA preservation or extraction methods or kits. Commercial buffers and kits such as those used in this study can be discontinued or the recipe can change (this was recently the case with the PowerSoil kit, which was discontinued as MoBio was taken over by Qiagen), thereby making specific recommendations meaningless within a short time.
However, in light of our results, we recommend that the extraction and preservation method should be chosen carefully depending on the specific groups of interest in the focal ecosystem. If soft-bodied microbes like nematodes and other microscopic metazoans are especially important to recover, gentle lysis methods such as chemical and enzymatic lysis may be preferred over harsh mechanical lysis.
Conversely, lysis of organisms with hard cell walls or frustules, such as diatoms, may benefit from mechanical methods such as bead beating.
Finally, although preservation via RNAlater does impact perceived community composition in both eukaryotes and prokaryotes significantly, it still offers representative community profiles and even appears to mitigate the effect of DNA extraction bias for eukaryotes.
Therefore, we recommend preservation in RNAlater (and other similar buffers) as a practical and adequate alternative to flash-freezing.

ACKNOWLEDGMENTS
The authors wish to thank Janina Brakel, Thorsten Reusch, and  Then, each plant was put in a plastic bag with a little water from the tank in it and kept cool (0°C to +4°C) until sample processing the following day.

A.1.2. Sample preparation and preservation
To relate to the total leaf surface, the leaf widths and lengths of each leaf of all the plants were measured. Then, the leaves were rinsed with sterile filtered seawater (pore size = 0.2 µm) and the biofilm attached to the leaves was rubbed off with a sterile cotton swab. For the flash-frozen samples, the biofilm material was suspended in sterile filtered seawater, aliquoted to 1.5 ml reaction tubes, centrifuged to pellets, frozen in liquid N 2 , and stored at −20°C. To preserve the DNA in RNAlater, the biofilm material was suspended in RNAlater, aliquoted to 1.5 ml reaction tubes, centrifuged into a pellet, and stored at 4°C until DNA extraction.    A.1.6. Sequence processing Clipped sequences (adaptor and primer sequence remains removed) were processed using the DADA2 package (Callahan et al., 2016) in R (version 1.2.0) (R Core Team, 2018). Briefly, sequences were truncated to 200 bp length, filtered (maxEE = 2, truncQ = 2), dereplicated, and error rates were estimated with the maximum possible error estimates from the data as the initial guess.
Sample sequences were inferred and paired reads were merged. To remove chimeric sequences the removeBimeraDenovo function was used. The resulting unique sequence variants (ASVs) were taxonomically classified using the lowest common ancestor approach implemented in CREST (Lanzen et al., 2012) based on the Silva database (Pruesse et al., 2007).

A.1.7. Statistical analysis
Statistical analysis was carried out in R (R Core Team, 2018) using functions from the vegan package (Oksanen, 2022)