Performance of extraction methods for extracellular DNA from sediments across marine habitats

Determining the distribution of organisms is the foundation to many ecological studies and theories, but sampling frequency in space and time often limits our ability to measure broad‐scale patterns. To overcome these limitations, ecologists strive to create more efficient methods to enhance sampling coverage. Using environmental DNA (eDNA) to detect macroorganisms within aquatic environments is a promising technique that could broaden the scale of ecological stud‐ ies measuring the distribution of animals, as it is more cost effective Received: 25 July 2019 | Revised: 3 October 2019 | Accepted: 11 October 2019 DOI: 10.1002/edn3.48

Environmental DNA is the genetic material left by organisms in their environment. Depending on the goals of a study, environmental DNA can be extracted from many environmental matrices, including water or sediment. DNA can persist for months to millennia in sediment (Turner, Uy, & Everhart, 2015;Willerslev et al., 2014), so that extracting eDNA from sediment potentially offers an unprecedented potential to detect organisms in the ecosystem currently and in the past. The persistence of DNA in sediment results from adsorption to sediment particles (Deere et al., 1996;Lorenz & Wackernagel, 1987). However, sediment characteristics likely affect DNA adsorption (Ogram, Sayler, Gustin, & Lewis, 1988), which in turn can affect the extraction of eDNA. Moreover, the extracted solution of DNA may also include PCR inhibitors, such as humic substances present in silt and clay, which have been shown to inhibit amplification of DNA (Buxton et al., 2017;Lekang, Thompson, & Troedsson, 2015;Schrader, Schielke, Ellerbroek, & Johne, 2012). Thus, sediment characteristics could alter metabarcoding results and may limit the feasibility to compare communities detected by eDNA from different environments.
Studies have looked at how different sediments affect the quality and quantity of extracted DNA, as well as the diversity of organisms identified by metabarcoding. Buxton (Buxton et al., 2017) found that top soil, with high organic content, reduced detection of a species because of PCR inhibition, as compared to sand. Other studies have found that both sediment type and extraction protocol can affect results from eDNA. For example, the amount of extracted DNA was consistent among extraction protocols for sand but not clay, while the community from cloning was not affected by extraction protocol in sand but was affected in clay (Lekang et al., 2015). In addition, sediment type (soil or stream sediment) and commercial extraction kits used affected the quality and quantity of extracted DNA (Hermans, Buckley, & Lear, 2018). Thus, both extraction protocol and sediment type can affect results from eDNA, potentially hindering the use of sediment eDNA as a source of information for species richness across ecosystems.
Our goal was to provide robust procedures for using sediment eDNA to quantify diversity across marine habitats and more specifically to assess how results from eDNA of macroorganisms are affected by sediment type sampled from three tropical marine habitats, coral reefs, seagrass meadows, and mangrove forests. These habitats are important for eDNA studies as they have important ecological and economic roles (Barbier et al., 2011;Costanza et al., 1997), and together account for much of the biodiversity in tropical marine habitats (Gratwicke & Speight, 2005;Gray, 1997). Unlike former studies, we standardized the DNA within sediment as opposed to measuring the DNA within natural sediments, given that the amount of DNA may vary within and among habitats. Specifically, we conducted two experiments to achieve this goal: (a) measured the amount of extracellular DNA extracted from sediments that had DNA removed prior to a controlled DNA addition; (b) quantified the number of sequenced reads in sediments after we added an equal amount of DNA of a species not occurring naturally in the habitats.
In addition, we also tested three extraction methods during the second experiment to see if extraction method could interact with sediment type to affect the amount of extracted DNA.

| Sample collection
At each sampling, five subsamples were taken at each location by inserting 10 ml modified plastic syringes, with the tips removed (21 mm diameter opening), 5 cm into the sediment. Subsamples were collected haphazardly within a 100 m 2 area, each separated by at least 2 m. The subsamples were kept on ice, returned to the laboratory, and the top 8 mm of the sediment from each subsample was collected by pushing the core with a plunger until the correct amount was above the syringe and "cut" with a sterile plastic knife. The five subsamples were pooled to make one sample (2.8 cm 3 of sediment), which was then kept in 50-ml falcon tubes at −80°C until DNA extraction. Nitrile gloves were worn during collection and processing to reduce contamination, and all sampling materials were sterilized before use by soaking them in 10% bleach for at least 30 min. Samples were collected from three different habitats: coral reefs, seagrass meadows, and mangrove forests. Mangrove sediment was collected from moist sediment just above the water level. Collecting samples within the coral reef and seagrass meadow necessitated snorkeling because subtidal sampling was not feasible from the boat because of damaging the habitat-forming species. Sampling protocol with syringe was chosen to minimize contamination from the snorkeler and all precautions were taken to reduce contamination, which included sampling from down current, and quickly collecting and capping the sample. In addition, steps were taken to reduce sample contamination including conducting extraction and pre-and post-PCR preparation in separate areas where surfaces and instruments were deeply decontaminated with bleach and UV light.

| DNA extraction
We used a commercial extraction method for extracting DNA from sediments with DNA removed and added two additional extraction methods for testing the extraction of novel DNA. All protocols targeted extracellular DNA. The first protocol (referred to here as MO BIO) utilized a phosphate buffer solution which was added to the sediment; then, this buffer was purified using the MO BIO, PowerSoil DNA isolation kit (Taberlet et al., 2012) (MoBio Laboratories, Inc.).
First, a phosphate buffer solution (1.97 g of NaH 2 PO 4 and 14.7 g of Na 2 HPO 4 per liter of sterile water) was added to sediment in a 50-ml falcon tube for a 1:1 volume ratio (Taberlet et al., 2012) and gently mixed for 30 min using a HulaMixer (Life Technologies). The mixed samples were then centrifuged at 10,000 g for 10 min, and the supernatant was transferred into a 15 ml falcon tube and stored at −80°C until further processing. Then, 750 µl of the phosphate buffer containing extracellular DNA was processed using a MO BIO, PowerSoil DNA isolation kit skipping the cell lysis steps (skipped steps 1-12). The second extraction protocol was based on Lever et al. (2015) and now referred to as the Lever extraction. The published protocol was followed using the extracellular extraction with the carbonate removal treatment. This protocol was chosen because it has a carbonate removal step which is not present in the MO BIO protocol. The third protocol was a combination of the two protocols previously described, adding the extracted DNA from the MO BIO protocol back to the sediment (sediment was already treated with phosphate buffer for the MO BIO protocol) before conducting the Lever extraction. This protocol, which extracted DNA twice from the sediment, was accessed to see if combining both protocols improved DNA extraction. An extraction blank (nuclease-free water) for both extraction methods was included through all steps.

| Experiment using sediments with DNA removed
Sediment samples were collected from a seagrass meadow a coral reef and a mangrove forest along the central Saudi Arabian Red Sea coast (see Table 1 for details). Basic characteristics of the sediment were noted (preponderance of mud or sand), and organic and inorganic matter were measured from sediment (not used for the DNA extraction) by the loss-on-ignition method (Luczak, Janquin, & Kupka, 1997). Briefly, the loss-on-ignition method consisted of heating 3 g of the sediment to 375°C for 3 hr, and the lost mass was measured as organic material. The sediment was then heated to 800°C for 12 hr and the lost mass was measured as inorganic material.
To remove existing DNA in the sediment samples, we autoclaved and bleached the samples before we added a known amount of DNA. First, 20 g of a sediment sample from each habitat was autoclaved at 121°C for 80 min. Three samples from each habitat (0.5 ml) from the autoclaved sample were then put into a sterile 2-ml tube.
Second, the sediment was washed with 0.5 ml of bleach two times.
To remove the bleach, the sediment was then washed with nuclease-free water three times. A separate sample of sediment was then run through the same extraction steps as described for extracellular DNA and analyzed with a Qubit 2.0 fluorometer (Quant-IT dsDNA High Sensitivity Assay kit; Invitrogen) to ensure that the DNA concentration was below detection limit. In agreement with a past study (Otte et al., 2018), initial tests found that DNA was still present after autoclaving (tested with Qubit 2.0 fluorometer), so the bleach step was added. Ultrapure Salmon Sperm (Oncorhynchus keta; 0.5 ml) with a concentration of 100 ng/µl (Invitrogen, USA) was added to the DNA-free sediment samples and mixed for two hours in a HulaMixer.
After mixing, tubes were centrifuged and the supernatant was separated into a different 2 ml tube. The DNA in both the sediment (0.5 ml) and supernatant (0.5 ml added to phosphate buffer) were extracted using the MO BIO protocol previously described and analyzed with Qubit 2.0 fluorometer to determine DNA concentrations.

| Sequencing experiment
Sediment samples were collected from three coral reefs: three seagrass meadows, and three mangrove forests along the Saudi Arabian coast of the Red Sea (Table 1). Sites were chosen to broaden the breadth of our findings with different sediment sources both within and among the different habitats. Salmon sperm DNA at a concentration of 10 ng/µl and volume of 1 ml was added per 1 ml of sediment sample. Three grams of sediment was mixed with the salmon DNA for 2 hr using a HulaMixer. The sediment was then divided into individual 1 g aliquots, and DNA was extracted from these aliquots using each of the three different extraction protocols.

| Sequences analysis
The DADA2  For statistical analysis of the first extraction test using sediment with the DNA removed, the amount of extracted DNA from the three different habitats was tested for heterogeneity with the Levene's test and for differences among the means with a linear model (LM). For the second extraction test, the difference in organic and inorganic matter was tested with a LM. The difference in salmon reads among sediment from the three habitats and extraction protocol was tested with a 2-way LM with the fixed factors and the interaction between them. Significant differences among the levels within factors were determined from nonoverlapping standard errors from the model output. Summary statistics were calculated using the ANOVA function from the car package (Fox & Weisberg, 2011) for all LMs.

| Experiment using sediments with DNA removed
The seagrass sediments were muddy, and the mangrove sediments were primarily mud with some sand, while the coral sediment was almost exclusively coarse carbonate sand. The organic content of the sediment was 2.3% dry weight (DW) from the coral reef and the seagrass meadow, and 1.2% DW from the mangrove forest. The in-

| D ISCUSS I ON
Using eDNA to monitor the presence of animals is an emerging technique that can enhance our understanding of the distribution of organisms, particularly when using DNA in marine sediments, that contain a fingerprint of animals in the ecosystems integrated over time. We found that the amount of extracellular DNA extracted from sediments that had DNA removed prior to a controlled DNA addition and that the number of reads from an exogenous species did not differ among sediments from different important tropical habitats.
These findings suggest that the extraction of eDNA from sediment was not affected by sediment characteristics associated with these habitats. Although other variables such as eDNA degradation could affect cross-habitat comparisons, such as more rapid DNA degradation in coral reefs compared to that in seagrass meadows because of higher UV radiation (Strickler, Fremier, & Goldberg, 2015), these findings suggest that eDNA could be a useful tool in detecting communities among different marine habitats.
Sediment characteristics can affect the extraction of DNA and PCR products. For example, DNA adsorbs to sediment which reduces degradation but could also affect extraction (Lekang et al., 2015;Lorenz & Wackernagel, 1987). The use of heat and bleach to remove DNA from the sediment prior to adding a known amount of DNA could have altered the properties of the sediment that are known to affect adsorption, such as ionic strength and pH (Ogram et al., 1988), and the results could be different to what would be found with "untreated" sediment. The second experiment was designed to negate the potential artifact of DNA removal, and this experiment sequenced DNA after an exogenous DNA spike to the sediment.
This addition experiment corroborated the findings of the first and found that the number of reads of the added DNA did not differ across sediments from different marine habitats.
Extraction protocols can also affect the quality and quantity of DNA extracted from sediment (Hermans et al., 2018;Lekang et al., 2015;Lever et al., 2015). Extraction methods are developed to both isolate DNA, which can be affected by adsorption to sediments, and remove contaminants that inhibit PCR such as The amount of salmon extracellular DNA extracted from sediment (MO BIO protocol) after being treated to remove existing DNA and adding 5 µl of salmon DNA. There were no differences in variances (Leven's Test, p = .133) or means (ANOVA, p = .529) among the sediment collected from different habitats. Boxplots show the median with the upper and lower quartiles, while the whiskers extend to the extreme data point but no more than 1.5 times the respective quartile. The mean is indicated by the circle humic substances present in organic-rich sediment (Buxton et al., 2017). We found that there was no difference in the number of salmon reads between the Lever extraction and the other two methods, but that the MO BIO extraction had more salmon reads than the combination extraction. In the study that developed the extraction protocol, the Lever extraction had greater DNA yields than the MO BIO when tested on different sediment types, which was attributed to high PO 4 concentrations and making the pH > 9 during initial protocol steps (Lever et al., 2015).  (Geraldi, Wahle, & Dunnington, 2009). In addition, some fish species were not consistently detected when comparing surveys by visual census, traps, and video camera, which may impart be from biases associated with habitat characteristics (Bacheler et al., 2017). There is a clear need to develop and test survey techniques that have consistent species detections among different locations regardless of abiotic and biotic factors. This is particularly pertinent in complex marine habitats, such as coral reefs, mangrove forests, and seagrass meadows, which support high biodiversity including many cryptic species that take shelter within the structure. Our findings suggest that eDNA offers a new tool that could improve ecologists' ability to quantify and compare biodiversity among marine habitats.

ACK N OWLED G M ENTS
This research was supported by the King Abdullah University of Science and Technology through baseline funding and funds by the Tarek Ahmed Juffali Research Chair in Red Sea Ecology. We thank A. Anton, P. Carrillo de Albornoz, and J. Martinez Ayala for helping to collect samples. All authors declare they have not conflict of interest.
F I G U R E 2 The number of reads (a) and reads after rarefaction (b) of salmon sequenced from DNA that was extracted using 3 different protocols from sediment from three different habitats. Boxplots show the median with the upper and lower quartiles, while the whiskers extend to the extreme data point but no more than 1.5 times the respective quartile. The mean is indicated by the circle. Extraction and PCR blanks are shown, as well as positive blanks with four different concentrations of salmon DNA (0.001, 0.01, 0.1, and 1.0 ng/µl) as shown by size of asterisks

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sequences from this study have been deposited at NCBI's SRA under project no. PRJNA579138 (https ://www.ncbi.nlm.nih.gov/ sra/PRJNA 579138).