Assessing three soil removal methods for environmental DNA analysis of mock forensic geology evidence

Soil is useful in criminal investigations as it is highly variable and readily transferred. Forensic geologists use several different techniques to removal soil from evidence prior to the analysis of inorganic components. There has been recent interest from the forensic science community to analyze environmental deoxyribonucleic acid (eDNA) associated with soil to augment existing forensic analyses. Notably however, limited research has been conducted to compare commonly used soil removal methods for downstream eDNA analysis. In this study, three soil removal methods were assessed: picking/scraping, sonication, and swabbing. Three mock evidence types (t‐shirts, boot soles, and trowels) were sampled in triplicate with each removal method (n = 27). Soil samples underwent DNA isolation, quantification, and amplification of four genomic barcode regions: 16S for bacteria, ITS1 for fungi, ITS2 for plants, and COI for arthropods. Amplicons were prepared into libraries for DNA sequencing on an Illumina® MiniSeq. DNA concentrations were highest in picked/scraped samples and were statistically significant compared with swabbed and sonicated samples. Amplicon sequence variants (ASVs) were identified, and removal methods had no impact on the recovery of the total number of target ASVs. Additionally, when assessing each sample in multidimensional space, picked/scraped samples tended to cluster separately from swabbed and sonicated samples. The soil core used a reference in this study also clustered with the picked/scraped samples, indicating that these samples may be more reflective of the communities collected from soil cores. Based on these data, we identified that picking/scraping is an acceptable soil removal method for eDNA analysis.


| INTRODUC TI ON
Soil is a valuable material when submitted to forensic crime laboratories as trace evidence because it is highly variable (e.g., color, mineralogy, and texture), easily transferred during a commission of a crime due to its microscopic size, and often overlooked as evidence by perpetrators [1][2][3][4].Forensic geologists that examine soil evidence can be asked to answer a number of questions; however, the most common questions are related to sample-to-sample comparisons (e.g., comparing soil from the suspect's shoes to the crime scene soil) and provenance (or sample origin; e.g., where in the United States did the soil originate?).Unlike most forensic disciplines, forensic geology examinations do not necessarily follow a strict standard operating procedure, as the nature of the samples can be quite varied [5].
Therefore, removal of soil from evidence can be dependent on the size of the substrate (e.g., vehicle, clothing, digging equipment, etc.), the type of substrate (e.g., metal, cloth, rubber, etc.), and the amount and type of soil adhered.For example, small amounts of soil could be cut from fabric, but if the entire item is covered in soil, it may be more feasible to scrape soil from the item [6].
Forensic geologists have identified several approaches for removing soil from evidence, including hand picking with forceps, scraping, the use of adhesive tape, vacuuming, swabbing, and washing [4,6,7].The Organization of Scientific Area Committees for Forensic Science Trace Evidence Subcommittee developed a standard (ASTM E3272-21) to guide the collection of soil for criminal investigations [8].Within this standard, it recommends in general using tweezers, forceps, spatulas, or palette knives to collect small amounts of soil from items.For instance, flat tools such as razor blades can be useful in collecting thin layers of soil from non-porous substrates (e.g., exterior of a vehicle) and spatulas may be used to scrape soil from porous substrates (e.g., fabric).In scientific papers that included the removal of soil from evidence type substrates, little justification is provided on why a method of soil removal was selected.Case reports and research articles have minimally described utilizing scraping [9], brushing [10,11], swabbing [12], or sonication [10] to remove soil from various substrates.One study compared three different soil removal methods (dry brushing, washing, and tape lifting) for downstream analysis of minerals using a QEMSCAN system, and reported that tape lifting recovered more particles than dry brushing and washing [13].Notably, no published studies to date have examined or compared soil removal approaches from evidence type substrates for downstream DNA analysis of soil.
Harnessing environmental DNA (eDNA) from diverse organisms associated with soil has garnered interest in the forensic science community, as the biological community can be sample specific and informative of origin.DNA metabarcoding is an approach that allows for the amplification of small, yet, informative regions of the genome to characterize biological communities.Research assessing the bacterial [14,15], plant [14,16,17], insect [16,17], fungal [17,18], and eukaryotic [14,17] communities within mock soil core evidence has highlighted promising outcomes for use in forensic geology casework.Other studies have incorporated realistic crime scene scenarios by including mock evidence in their characterization of the bacterial [19,20] and eukaryotic [21] communities.In these studies, soil was either scraped from mock evidence [19,20], cut from articles of clothing [19,20], or not described as to how soil was removed [21].Given the many options to remove soil from evidence, the rationale for using scraping and cutting was not provided.To address an existing gap, the goal of this study was to compare three commonly used soil removal methods (swabbing, picking/scraping, and sonication) for removing soil from diverse mock evidence items to determine which was most appropriate for eDNA analysis.

| Study site setup
For this study, three t-shirts (Hanes Boys' Undershirt, EcoSmart Short Sleeve Crew Shirt, Winston-Salem, NC, USA), three rubber boot soles (KANEIJI Shoe Repair Rubber Full Sole, Yiwu City, Zhe Yiwu, China), and three metal trowels (Fiskars Ergo Gardening Hand Trowel, Keilaniemi, Espoo, Finland) were used, given they are commonly submitted evidence items.Polyurethane swabs (BD BBL™ CultureSwab™ EZ, Becton Dickinson, Franklin Lakes, NJ, USA) were used to sample the baseline biological community of each mock evidence before interaction with soil.A 122 cm long and 76 cm-wide area at the study location (Raleigh, NC, USA) was removed of all grass and clover.The top layer of soil (~5 cm) was collected into a plastic bin and homogenized using the three trowels.Soil was subsequently rubbed into the trowel blades (not the handles), the patterned/impression side of the boot soles, and the front side of each t-shirt.The soil was damp from previous precipitation, making it possible for the soil to easily adhere to mock evidence.A 50 mL conical tube was filled with soil from the bin to be used as a reference and allowed to dry over night at room temperature in a fume hood.Once soil was adhered to the mock evidence items, the remaining soil in the bin was placed back into the area it was originally collected from and items were placed to make contact with the area as follows: trowels on top of the soil, soles pattern/impression side down touching the soil, and t-shirts front side down.The shirts and boot soles were staked into the ground to ensure minimal movement from animals and wind.A 122 cm by 76 cm wire dog crate (New World Pet Products, Muncie, IN, USA) was placed over the area containing mock evidence to further protect the samples.The trowel handles were zip-tied to the crate and the crate was also staked into the ground to reduce movement.Mock evidence was left outdoors for 2 weeks before soil removal methods were tested.

| Soil removal methods
For each removal method, one t-shirt, one boot sole, and one trowel were collected and sampled in triplicate (Figure S1).When using the swabbing method, BD BBL™ CultureSwab™ EZ swabs were moistened with 20 μL of sterile molecular grade water and each item was swabbed in a unique area (Figure S1) until the swab head was saturated with soil.After sampling, swabs were placed back into their dry transport containers until they could undergo DNA isolation.To collect soil via picking/scraping, mock evidence was placed on a clean sheet of butcher paper and allowed to dry overnight.
Using sterile forceps and scoopula, soil was subsequently scraped onto small pieces of butcher paper.Once soil was scraped off, soil was poured into 5 mL tubes for temporary room temperature storage prior to DNA isolation.A sonicating bath (Fisherbrand™ CPXH Series Heated Ultrasonic Cleaning Bath, Fisher Scientific, Hampton, NH, USA) with a tray insert (Branson Ultrasonics, Brookfield, CT, USA) were used for separately sonicating one of each mock evidence type at max power (40 kHz) for 10 min.The t-shirt was placed into a plastic container and covered with 10% ethanol.The plastic container was placed inside the sterilized tray (which was filled with water) and sonicated.The original plastic container caused the tshirt to not have much room, so a wider plastic container was used and sonicated again using the same parameters.The boot sole was cut into thirds to fit into a separate plastic container.Enough 10% ethanol was added to cover the three boot sole pieces and they were sonicated together.Lastly, the trowel was placed into a third plastic container, but was only half submerged in 10% ethanol and additional autoclaved water was added to fully submerge the item given its size.The sonicated contents and liquids for each of the three items were each carefully transferred into separate 1 L sterilized glass bottles and 50 mL conical tubes for storage.The liquid and soil were gently shaken and 250 mL of each sonicated liquid were filtered separately using 45 micron mixed cellulose ester filters (MilliporeSigma, Burlington, MA, USA) with the MilliporeSigma Classic Glass Vacuum Filter Holder Kit.Given the poor vacuum pressure and the large amount of soil that adhered to the filters, only ~100 mL of liquid was filtered for both the boot sole and t-shirt and ~125 mL for the trowel.The filters were gently removed to petri dishes to dry overnight at room temperature.Each of the three filters were subsampled by cutting into 1 cm 2 pieces.

| DNA isolation and quantification
DNA was isolated using the PowerSoil Pro Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol with two modifications: (1) after the addition of CD1 in step one, 20 μL (20 mg/mL) of Proteinase K (Qiagen) was added to the sample and allowed to incubate for 30 min at 65°C, and (2) DNA was eluted into two 50 μL eluates [22].Input for DNA isolations were as follows: Baseline and swabbing -swab heads were cut directly into the sample isolation tubes; Scraping/Picking and Reference Core -Plastic disposable scoops (VWR International, Radnor, PA, USA) were used to transfer ~100 mg of soil from scraped samples; Sonication -three of the 1 cm 2 subsamples cut from each filter were placed directly into the soil isolation tubes.Each removal method had nine samples that underwent DNA isolation, three for each mock evidence substrate.DNA was quantified from both eluates with the dsDNA HS kit (Invitrogen, Waltham, MA, USA) using the Qubit Fluorometer (Invitrogen) following the manufacturer's recommendations and stored at −20°C when not in use [23].

| DNA amplification, pooling, purification, and quantification
Two replicates from each mock evidence item and removal method (total n = 18), three controls (one reagent blank, one positive control, and one negative control), three baseline samples, and one soil reference were amplified in duplicate for each primer pair (Table S1) targeting 16S, ITS1, ITS2, and COI using the KAPA3G Plant PCR Kit (Roche, Basel, Switzerland).Each 12.5 μL reaction contained 6.25 μL of 2X KAPA Plant PCR Buffer, 0.1 μL of KAPA3G Plant DNA Polymerase, 0.3 μM of each primer and 2 μL of DNA (or 2 μL of controls).Additionally, for the ITS2F/p4 primer pair, DMSO was added to the master mix (final concentration 4%).
Following PCR, all amplicons were combined for each replicate for a final volume of 100 μL.A ratio of 1.4X of AMPure XP Beads (Beckman Coulter, Brea, CA, USA) to sample was used for amplicon purification following the manufacturer's recommendations [24].Amplicons were eluted from the beads using 60 μL of Buffer EB (Qiagen).Purified amplicons were quantified using the dsDNA HS kit (Invitrogen) and the Qubit Fluorometer (Invitrogen) using 2 μL of sample [23].

| Library preparation, quantification, pooling, and sequencing
The KAPA HyperPlus Kit (Roche) was used to prepare libraries from 50 μL purified amplicons with Illumina ® compatible indices and adapters [25].The manufacturer's protocol was followed with the exception of (1) not performing the fragmentation step, and (2) an incubation step of 20°C for 30 min prior to the end repair and A-tailing thermocycler setting of 65°C for 30 min.Libraries were eluted in 25 μL of Buffer EB (Qiagen).Libraries were quantified via qPCR on the QuantStudio5 (Applied Biosystems, Waltham, MA, USA) with the KAPA Library Quantification Complete kit (universal; Roche).Libraries were diluted 1:10,000 for quantification and quantified following the manufacturer's instructions [26].Each library (total n = 25) was pooled in equimolar concentrations and the library pool was quantified via qPCR (as described above) and adjusted for the average size of library (~500 base pairs).The pool was diluted to a final loading concentration of 1.4 pM following the manufacturer's guidelines with PhiX spiked-in at 20% [27].The pool was sequenced on an Illumina ® MiniSeq™ (Illumina ® , San Diego, CA, USA) using the Mid Output Kit (1 × 300 bp).

| Bioinformatics
A bioinformatic pipeline that was developed for the Meiklejohn lab was used to analyze raw sequence data to determine the amplicon sequence variants (ASVs) recovered from each sample using NC State's High Performance Computing Services.Briefly, raw data was fed into cutadapt [32] to trim primer and adapter sequences from reads and DADA2 [33] was employed to identify ASVs and determine their abundance (based on read depth).Following ASV classification, each ASV was queried against GenBank's nucleotide database via BLAST [34] generating a list of taxids.These taxids were then taxonomically identified using taxizedb [35].For each ASV, the results were filtered to the top 10 hits based on an e-value ≤0.001, a percent identity of ≥95%, and a percent query coverage of ≥90%.The results were further filtered to select the top hit(s) based on the maximum percent identity.[36] utilizing the vegan (version 2.6-4) [37] and ggplot2 (version 3.4.2) [38] packages using the default parameters with a Bray-Curtis distance matrix.

| Data analysis
For these analyses, the positive control and baseline samples were not included.

| RE SULTS AND D ISCUSS I ON
DNA yields were determined via Qubit for each mock evidence item and soil removal method (Table 1).Scraping/picking of soil from mock evidence items resulted in significantly higher amounts of DNA compared with swabbing (p = 0.002) and sonication (p = 0.007).
Notably, swabbing was the easiest method to complete, but once swab heads were saturated with soil, no more material was able to be collected.Sonication was the most difficult to implement as it required multiple containers and volumes of 10% ethanol for submersion.Additionally, the filtering process was very labor-intensive, and filters quickly became clogged with material.Picking/scraping of soil was the second easiest and quickest approach for the removal soil from mock evidence, with most of the soil easily removed from mock evidence.Based on the differences in DNA yield, as well as feasibility, picking/scraping was preliminarily selected as the method of choice for soil removal.
Considering the broad range of DNA yields from each method, it was necessary to determine if the biological communities recovered were dependent on the soil removal method.DNA metabarcoding of 16S, ITS1, ITS2, and COI was completed to assess the bacterial, fungal, plant, and arthropod communities, respectively, obtained from mock evidence using the three different removal methods.ASVs were identified for each method, mock evidence item, and genomic barcode region (Figure 1) and subsequently categorized as either target (e.g., 16S sequences correctly identified as bacteria), non-target (e.g., 16S sequences incorrectly identified as fungi), or unclassified.The results in Figure 1 demonstrate the high specificity of primer pairs selected for bacteria, plants, and fungi.However, the majority of reads were unclassified for arthropods, suggesting that the database may not contain the sequences for the arthropod taxa that were recovered.Statistical analyses indicated that there was no significance between (a) the number of target ASVs and soil removal method (F Ratio = 0.0279; p = 0.9996), and (b) the number of target ASVs and mock evidence item (F Ratio = 0.0264; p = 0.9739).
Taxon abundance charts were generated using only the target ASVs recovered from each sample to visualize the bacterial, fungal, plant, and arthropod communities, along with their relative proportions (Figures S2-S5).Taxon abundance charts provide community composition in simple visualizations to glean preliminary information and to help assist in next steps for analysis.To simplify interpretation, fungi, plants, and arthropods were plotted down to the family level, whereas bacteria were plotted at the order level.Additionally, NMDS plots were created to determine how replicates differed from each other in multidimensional space (Figure 2).For bacteria (Figure 2A), sonicated and swabbed replicates clustered together whereas the picked/scraped replicates clustered separately.Additionally, regardless of removal method, distinct groupings by mock evidence were also observed for trowel and boot sole samples, and most of the t-shirt samples grouped together.When looking at fungi (Figure 2B), two trends were observed: (1) samples were grouped in multi-dimensional space mostly based on mock evidence type (with t-shirt samples most disparate), and (2) sonicated replicates had the clearest clustering compared with the other removal methods.Plant taxa recovered using picking/scraping grouped together and all replicates from swabbing and sonication were clustered together (Figure 2C).
Arthropods have only a single cluster containing the sonicated replicates and no clear clustering pattern for the other replicates (Figure 2D).When looking at all four taxa combined, three distinct clusters are apparent -one containing most of the swabbed replicates, one with all of the sonicated replicates, and the other containing most of the picked/scraped replicates (Figure 2E).Of note, the soil reference that was also collected from the study site clustered with the picked/scraped replicates when looking at TA B L E 1 Average DNA concentration and standard deviation across triplicate samples for each soil removal method and mock evidence item.taxa combined (Figure 2E).This suggests that the picking/scraping method most reflects the communities found in ~100 mg scoops of soil compared with swabbing and sonication.

Swabbing
This study is the first of its kind to compare different soil removal strategies for eDNA analysis of mock forensic evidence.Each method yielded sufficient DNA for analysis and there were no statistically significant differences in the number of target ASVs recovered.Based on the NMDS plots, it is evident that there are compositional differences between picking/scraping compared with swabbing and sonication.Considering these data, and determining the ease at which these removal methods can be completed, we found that picking/ scraping was most appropriate for eDNA analysis using soil collected from mock evidence.

| CON CLUS IONS
Forensic scientists have described several ways to remove soil from evidence; however, there has not been a study that has determined which soil removal method is most applicable for downstream eDNA analysis.The commonly used methods of swabbing, sonication, and picking/scraping were selected to assess both overall DNA yields and compare the bacterial, fungal, plant, and arthropod communities recovered from mock evidence.Picking/ scraping of t-shirts, boot soles, and trowels produced the highest DNA concentrations compared with the other two methods.
ASV classification from DNA metabarcoding of the 16S, ITS1, ITS2, and COI genomic barcode regions found that none of the removal methods outperformed each other when assessing the recovered target ASVs.However, for taxonomic assignment, replicates from picking/scraping method tended to cluster together whereas swabbed and sonicated replicates were either grouped together or clustered separately from each other depending on genomic barcode region.Importantly, the soil core used as a reference clustered with the picked/scraped replicates, indicating that picking/scraping may more closely resemble the baseline soil compared with the other two approaches.As for feasibility, swabbing was the easiest method followed by picking/scraping then sonication.Given the ease of picking/scraping, as well as the high DNA yields, no statistical difference in target ASV classification, and distinct clustering of picked/scraped replicates, we found that picking/scraping was the most appropriate method for soil removal for eDNA analysis in forensic geology casework situations.
A student's t-test was applied to determine statistical significance between DNA yields from each soil removal method in Microsoft Excel (Redmond, WA, USA).Taxon abundance charts were generated in Tableau (version 2022.4;Mountain View, CA, USA).JMP Pro (version 15; SAS Institute, Cary, NC, USA) was used to assess statistical significance of each method using a one-way ANOVA and applying the Tukey-Kramer HSD test.Lastly, non-metric multidimensional scaling (NMDS) plots were created in RStudio (version 2023.030+386[R version 4.3.0]) individually (Figure 2A-D) and when viewing all four

Future work exploring other
types of evidence (e.g., tires, polyester clothing, rope, etc.) with these removal methods are needed to better understand if picking/scraping can be a universal method for eDNA analysis.Notably, this study only focused on surface soil, which is the most common type of soil submitted to crime labs.While conclusions drawn on removal methods are unlikely to change if a different horizon of soil was submitted for analysis, additional work comparing removal methods with varied soil depths could be valuable.Additionally, this research only utilized a clay-rich soil, thus research assessing these removal methods F I G U R E 1 Soil removal methods and amplicon sequence variant (ASV) Assessment.The figure displays the total number of target, nontarget, and unclassified ASVs of both replicates for each mock evidence item (t-shirt, boot sole, and trowel), each genomic barcode region (16S [bacteria], COI [arthropods], ITS1 [fungi], and ITS2 [plants]), and removal method (picking/scraping [labeled "scrape"], sonication, and swabbing [labeled "swab"]).F I G U R E 2 Non-metric multidimensional scaling (NMDS) plots comparing soil removal methods.NMDS plots were generated from target amplicon sequence variants (ASVs) that were taxonomically identified.Soil removal methods are indicated by color, with green representing the soil core used for a reference, and mock evidence denoted by shape.The bacteria [16S] (A), fungi [ITS1] (B), plants [ITS2] (C), arthropods [COI] (D), and all four taxa combined (E) are displayed.