Evaluation of an automated protocol for efficient and reliable DNA extraction of dietary samples

Abstract Molecular techniques have become an important tool to empirically assess feeding interactions. The increased usage of next‐generation sequencing approaches has stressed the need of fast DNA extraction that does not compromise DNA quality. Dietary samples here pose a particular challenge, as these demand high‐quality DNA extraction procedures for obtaining the minute quantities of short‐fragmented food DNA. Automatic high‐throughput procedures significantly decrease time and costs and allow for standardization of extracting total DNA. However, these approaches have not yet been evaluated for dietary samples. We tested the efficiency of an automatic DNA extraction platform and a traditional CTAB protocol, employing a variety of dietary samples including invertebrate whole‐body extracts as well as invertebrate and vertebrate gut content samples and feces. Extraction efficacy was quantified using the proportions of successful PCR amplifications of both total and prey DNA, and cost was estimated in terms of time and material expense. For extraction of total DNA, the automated platform performed better for both invertebrate and vertebrate samples. This was also true for prey detection in vertebrate samples. For the dietary analysis in invertebrates, there is still room for improvement when using the high‐throughput system for optimal DNA yields. Overall, the automated DNA extraction system turned out as a promising alternative to labor‐intensive, low‐throughput manual extraction methods such as CTAB. It is opening up the opportunity for an extensive use of this cost‐efficient and innovative methodology at low contamination risk also in trophic ecology.

Examining these feeding networks with molecular methods usually entails processing large numbers of samples: Raso et al. (2014), for example, tested more than 2,500 invertebrate predator samples for extra-and intraguild prey, and Gariepy, Kuhlmann, Gillott, and Erlandson (2008) screened DNA extracts of 26,000 field-collected mirid host samples for the occurence of parasitoid DNA. The use of low sample numbers entails the risk of only capturing a smaller proportion of the actually consumed prey species (Burgar et al., 2014) and/or missing statistically significant differences in the trophic variables examined.
Accordingly, high sample numbers, which are representative for the trophic interactions studied, need to be analyzed molecularly.
Here, DNA extraction often represents a bottleneck as, depending on the protocol used, it can be tedious, time-consuming, and expensive. Moreover, the presence of contaminating and potentially PCR inhibiting substances can impede many of the subsequent reactions and techniques (Berensmeier, 2006;Zarzoso-Lacoste, Corse, & Vidal, 2013). Accordingly, efficient DNA extraction protocols are needed to generate high-quality DNA from various types of complex dietary samples that often only contain trace amounts of food DNA (King et al., 2008;Pompanon et al., 2012). Among the conventional extraction methods, a range of approaches is known for DNA isolation in the fluid phase. They involve lysis by a detergent or chaotropic substance (possibly in the presence of protein-degrading enzymes), followed by several processing steps applying organic solvents such as phenol and/or chloroform or ethanol, which in general are toxic and require special and expensive disposal. Alternative separation techniques are employing solid-phase systems, where sorption processes are based on silica under chaotropic conditions, ionic exchanges, as well as affinity and size exclusion mechanisms (for a review, see Berensmeier, 2006). These classical DNA extraction methods are not only laborsome and time-consuming, but the relatively large number of steps involved increases the risk of DNA degradation, sample loss, or DNA cross-contamination, especially when high numbers of samples need to be processed simultaneously.
During the last few years, paramagnetic beads have been increasingly employed for DNA isolation, representing a relatively easy and inexpensive technology, subjecting samples to very little mechanical stress (Berensmeier, 2006;Suomalainen, Suomalainen, Puro, Kytöniemi, & Lamberg, 2010;Vidergar, Toplak, & Kuntner, 2014 reagents are moved into and out of a single well to perform the different steps of a DNA isolation procedure, permanent magnetic rods are used collecting paramagnetic beads to which the DNA is bound and released into 96-well plates (Fang et al., 2007). Automated extractions obtain purified DNA in sufficient quality and purity and proved to be consistent and reproducible (Loeffler, Schmidt, Hebart, & Einsele, 2004). Centrifugation steps are avoided and sample handling steps are reduced, and thus, the risk of cross-contamination is lowered. Compared to classical DNA extraction methods, the magnetic separation of DNA has several advantages: The simplified procedures employing a robotic workstation for magnetic particle handling entail increased worker safety together with reduced sample processing time leading to increased laboratory efficiency (Boyd, 2002). The method produces good yields of high-purity DNA appropriate for a variety of downstream applications (Fang et al., 2007;Tan & Yiap, 2009;Wochner, Birgit Cech, Menger, Erdmann, & Glökler, 2007) and hence might be especially suited for large-scale studies (Berensmeier, 2006).
In this study, we tested a high-throughput DNA extraction platform using paramagnetic beads for the detection of prey DNA in dietary samples from whole-body invertebrate samples, vertebrate gut contents, and feces and provided an evaluation of its performance in comparison with a classical, well-proven CTAB-based DNA extraction protocol. Extraction efficacy was quantified using the proportions of successful PCR amplifications of both total and prey DNA, and cost was estimated in terms of time and material expense.

| MATERIALS AND METHODS
Two different DNA extraction methods were compared, namely a CTAB-based phenol-chloroform protocol ("CTAB samples") and a silica-based automated DNA extraction ("BioSprint samples") on the BioSprint ® 96 extraction platform using the BioSprint ® 96 DNA Blood & Tissue Kit (Qiagen, Hilden, Germany). We selected different kinds of dietary samples to cover a broad variety of sample types, that is, gut content, faces, as well as whole-body extracts, from different representatives of both invertebrates and vertebrates: 25 predatory arthropods (carabid beetles and spiders) from cereal fields, 52 ladybird beetles from a feeding experiment, 15 omnivorous carabid beetles of the genus Amara sp. from a glacier foreland,

| Lysis and DNA extraction
For the field-collected predatory arthropods, the ladybird beetles from the feeding experiment, the insect omnivorous Amara sp. specimen, and the plant samples, the lysis was conducted as follows: Whole specimens and plant tissue, respectively, were put individually in 2-ml reaction tubes adding 430 μl 1 × TES buffer (0.1 mol/L TRIS, 10 mmol/L EDTA, 2% SDS; pH 8), 10 μl proteinase K (20 mg/ml), and 5-8 glass beads (Ø 3 mm) each. After grinding in a Precellys ® 24 Tissue Homogenizer (Bertin Technologies, Montignyle-Bretonneux, France) at 21,000 g for 2 × 60 s, they were incubated at 58°C for 24 hr. For plants and Amara sp., an additional 1 mg PVP (polyvinylpyrrolidone) was added to the lysis buffer to remove PCRinhibiting phenolic compounds stemming from plants. The vertebrate samples (stomach/gut content of cormorants, fecal pellets of salamanders) were treated similarly, without the grinding step, whereas for the cormorant samples, we raised the volume of 1 × TES buffer and proteinase K to 980 μl and 20 μl, respectively, due to their bigger initial volume. Per lysate, 200 μl each was used for CTAB and BioSprint DNA extraction.
The CTAB extraction was performed as follows: 60 μl of 5 mol/L NaCl and 25 μl of 10% CTAB (cetyltrimethylammonium bromide) solution were added to the 200 μl lysates and incubated for 10 min at 65°C. Next, 300 μl of chloroform:isoamylalcohol (24:1) was added, and the samples were allowed to rest for 10 min before they were centrifuged for 5 min at 18,000 g, and the aqueous layer transferred into fresh reaction tubes. After adding 100 μl of 5 mol/L NH 4 Ac, samples were placed on ice for at least 30 min and centrifuged for 20 min at 4°C 21,000 g, and the liquid phase was transferred into a fresh tube.
DNA was precipitated with the same volume of isopropanol (approx.

| PCR, visualization, and statistical analysis
CTAB and BioSprint samples were initially tested in a PCR using general primers to check whether they contain amplifiable DNA and then with prey-specific primers to specifically assess the detection of prey DNA in the samples (primer details and the respective PCR conditions are provided in Data S2). PCR products were visualized using QIAxcel ® , an automated capillary electrophoresis system (Qiagen), with method AL320 on the QIAxcel Screening Kit, and the results were scored with BioCalculator Fast Analysis Software version 3.0 (Qiagen). Samples showing the expected fragment length with a signal above 0.05 and 0.1 relative fluorescent units were deemed positive, depending on the PCR assay.
To compare the performance of the different extraction methods, we used the PCR amplification success with both general and prey-specific primers. Dietary samples are a mixture of only minute amounts of prey DNA together with high concentrations of predator DNA (King et al., 2008). Therefore, a quantification of prey DNA yield via devices such as Nanodrop ® or Qubit ® is impossible. Differences in DNA detectability between CTAB and BioSprint samples were statistically compared by chi-square tests. All tests were carried out with SPSS 21 (IBM, Armonk, NY, USA).

| Expenditure of time and money
Costs for DNA extraction per sample (excluding the lysis step) were estimated in terms of time and material expenses. Material affordability was calculated based on the list price for necessary supplies and reagents (as of January 2017). Start-up costs for the BioSprint ® 96 as well as standard laboratory equipment such as centrifuges, thermoblock, laminar flow, and extraction hood were excluded. Also the payment of the work was excluded as it is strongly dependent on the employment contract of the executing person.

| Extraction efficacy via PCR amplification success
For general primers, PCR amplification success was high across all samples ( Figure 1). Here, the BioSprint extraction method generally turned out to be more effective in isolating amplifiable DNA than CTAB, as its performance was significantly higher for both groups together, vertebrates and invertebrates (χ 2 = 10.9, p < .001). Among those samples which tested positive (i.e., successful PCR amplification) with only one of the two DNA extraction methods, the BioSprint extraction method had six times higher PCR amplification success than the CTAB extraction ( Figure 1). This effect is mainly rooted in the large difference in amplification success for the vertebrate samples (stomach/gut content of cormorants and salamander feces) between BioSprint and CTAB (81.1% vs. 37.7%; χ 2 = 20.7, p < .001). For the invertebrate samples, there was no such difference present (87.2% vs. 88.9% n.s.). All plant and ladybird beetle samples tested positive with both extraction methods using general primers (Table 1).
With the prey-specific primers, prey DNA detection rates for all samples together were similar for both kinds of extraction methods: CTAB 45.6% and BioSprint extraction 44.4%, respectively (Figure 1). However, the situation changed when considering vertebrates and invertebrates separately: in vertebrate samples, the amplification success of prey DNA was significantly higher for BioSprint samples (73.6%) than for CTAB samples (47.2%, p < .005). Contrastingly, for the invertebrate samples, less BioSprint samples tested positive for prey DNA (33.3%) compared to the CTAB samples (45.3%, p = .061; Figure 1).
Overall, there was no group of invertebrates where BioSprint samples had a better performance than CTAB samples, whereas for vertebrates the opposite was true (Table 1).
Among the invertebrates, there were no Amara sp. samples with amplifiable DNA exclusively for BioSprint (i.e., they either were positive exclusively for CTAB or with both extraction methods), neither with the general nor with the prey-specific primers. Contrastingly, among the vertebrates, no cormorant stomach/gut content samples tested positive exclusively for the CTAB extraction.

| Expenditure of time and money
Material costs per sample were one-third cheaper for CTAB DNA extraction than for the BioSprint ® 96 extraction platform using the BioSprint ® DNA Blood and Tissue Kit (Table 2). Yet, the time expenditure for CTAB samples was over eight times higher than for BioSprint samples. The CTAB extraction implies two transfers per sample of parts of the lysate from one reaction tube to a fresh one. Moreover, reaction tubes need to be opened and closed six times during the procedure in order to add reagents. Contrastingly, the BioSprint samples are only opened once, that is, when the lysate is transferred into the 96-well plate going into the extraction device. This could lower the risk of carry-over contamination.
F I G U R E 1 DNA amplification rates in invertebrate (n = 117) and vertebrate samples (n = 53) extracted with CTAB protocol and the BioSprint ® platform together with the DNA Blood Kit using general primers (left) and prey-specific primers (right); for all samples taken together and separately for invertebrates and vertebrates, respectively. Amplification success for plant samples was 100% for both extraction methods (data not shown). CTAB only represents the share of samples that tested positive when they were CTAB-extracted and negative when using BioSprint ® 96. For BioSprint only, this was exactly the other way round, that is, the share of samples that tested positive for BioSprint ® 96 and negative for CTAB. Both positive is the share of samples with successful PCR amplifications for both extraction methods. Asterisks indicate significant differences

| DISCUSSION
We compared the performance of a high-throughput DNA extraction procedure, with the example of the BioSprint ® 96 extraction platform in combination with the commercial BioSprint ® 96 DNA Blood and Tissue Kit, and a classical, well-proven CTAB-based protocol for the detection of prey DNA in dietary samples from whole-body invertebrate samples, vertebrate stomach/gut contents, and feces. Using general primers, the two extraction methods resulted in similar rates of amplifiable total DNA, meaning that the PCR amplification success for total DNA was over 90% in both cases. Here, the performance of the BioSprint samples was higher for both vertebrates and invertebrates.
This indicates that automated magnetic separators are more effective in extracting total DNA than the CTAB-based protocol. Moreover, automatic DNA extraction procedures have the huge advantage that they generally significantly decrease time and costs. Ivanova, DeWaard, and Hebert (2006), Ivanova, Fazekas, and Hebert (2008), for example, optimized a semiautomated DNA extraction method for animal tissue and plants using glass fiber 96-well plates (PALL Inc.) in combination with a Biomek ® FX liquid handling station (Beckman Coulter). The performance of this method was comparable to singletube commercial DNA isolation kits. However, Ivanova's approach still implies numerous steps of handling with plates and chemicals during the extraction process, whereas devices such as BioSprint ® are designed for fully automated DNA extraction once loaded with lysate and buffers. Because of this, the risk of contamination is significantly reduced, allowing for standardized DNA extraction procedures. This can be seen, for example, in high-throughput protocols using paramagnetic separation techniques in combination with automated DNA extraction systems for detecting enteric livestock diseases in feces (Chen et al., 2014;Plain et al., 2014). Additional benefits of the platebased DNA extraction of automated platforms in comparison with individual tube-based extractions are the overall reduction in plastic waste and the usually smaller reagent volumes required (Schiebelhut, Abboud, Gómez Daglio, Swift, & Dawson, 2016).
Automated DNA extraction has not yet been evaluated for dietary samples. The present results indicate significant differences between the two extraction methods when using prey-specific primers for vertebrates and invertebrates, with higher prey DNA detection rates for BioSprint samples in vertebrates. The results are considered to be representative for any of the platforms using this well-proven principle T A B L E 1 Overview of the DNA detection success of the samples comprising different vertebrate, invertebrate, and plant taxa. (i) Invertebrates, (v) vertebrates; targeted genes: COI (mtDNA), 16S (mtDNA), CDH1 (nDNA), 18S (nDNA), and trnL (cpDNA). CTAB only (%) represents the share of samples that tested positive when they were CTAB-extracted and negative when using BioSprint ® 96. For BioSprint only (%), this was exactly the other way round. Both positive (%) is the share of samples with successful PCR amplifications for both extraction methods and Both negative (%) is those which never delivered a PCR product   (Jarman et al., 2013), where a "Maxwell 16" DNA extraction robot (Promega) was used, also working with paramagnetic beads. The contrasting higher recovery rate of prey DNA in CTAB-compared to BioSprint samples of invertebrates may be rooted in the fact that the former extraction protocol represents a constantly improved procedure of working steps, where most of the optimization was performed in regard to the detection of prey DNA in whole-body extracts of invertebrates (Juen & Traugott, 2005;Oehm et al., 2011;Raso et al., 2014;Wallinger et al., 2013). We kept on continuously adapting and optimizing these CTAB protocols depending on our needs for different predator and "prey"  (King et al., 2008). In vertebrate feces, a considerable fraction of DNA can originate from cells of the intestinal mucosa of the defecating animal, too (Albaugh et al. 1992;Deagle et al. 2006). However, the share of consumer DNA is far below the one in whole-body invertebrate extracts. The weaker performance of BioSprint ® 96 for prey DNA detection in invertebrates might be attributable to the fact that the silica surface of the paramagnetic beads is fully occupied by consumer DNA, so that there is no more binding capacity left for prey DNA. A potential solution here would be the use of invertebrate regurgitates (Waldner & Traugott, 2012;Wallinger et al., 2015) or feces   of extraction methods during studies should be avoided.
One of the biggest advantages of employing robotic DNA extraction platforms is that they save time as they speed up the analysis and reduce analytical error. Compared to classical DNA extraction approaches, their use in combination with commercial kits allows to run up to 96 samples at once with minimal hands-on time (Carter et al., 2010). Albeit such automated systems are not cheap, it is paying off in a rather short time at a corresponding use of capacities and eventually a joint usage among different labs. Furthermore, automated platforms reduce the risk for cross-contamination due to the few manual handling steps compared to CTAB-or column-based silica extraction protocols. Among these platforms, BioSprint ® 96 or Kingfisher ® 96 offer the advantage of being "open systems" allowing tailoring of one's own methods. Although predefined application protocols are available, the results of the present study indicate the relevance of an option for customizing existing protocols or creating entirely new ones to meeting specific requirements. For example, different mixing speeds turned out to have a strong effect on the DNA purification (Suomalainen et al., 2010).
In conclusion, paramagnetic particle-based DNA purification systems, employing an automated platform such as BioSprint ® 96, achieve high rates of amplification success for total DNA form a variety of samples. Moreover, the method proved to be rapid, efficient, and reliable for DNA extraction also in different types of dietary samples. The system tested here is highly recommendable for large-scale dietary studies. The drastically reduced time effort and the low risk of cross-contamination play a critical role in high-throughput analysis. The system has a number of advantages regarding laboratory use: Parallelization allows for simultaneous handling of different targets and increased throughput. Moreover, the risk of cross-contamination is drastically reduced. Although there is still scope for improvement, the present results demonstrate that paramagnetic bead-based DNA extraction in combination with the use of automated extraction platforms is a promising tool for investigating trophic interactions comprehensively and is highly recommendable for high-throughput analysis.

DATA ACCESSIBILITY
The data of field experiment and molecular analysis will be uploaded to DRYAD upon publication.