Drill core and seawater samples were collected during IODP Exp. 357 and are described in detail in Motamedi et al. (2020). Briefly, core subsamples identified for microbiological analyses were immediately retrieved from the drill upon its return to the ship deck, wrapped in acid-washed and autoclaved teflon, and stored at -80 ∘C. Core samples were then shipped to the Kochi Core Center (Japan) for further processing and subsampling under sterile conditions (Früh-Green et al., 2017b; Orcutt et al., 2017). Please see Früh-Green et al. (2017a, 2018) for more details on lithology and other details of the recovered cores.

Prior to the deployment of the drill at each site, the ship's conductivity, temperature, and depth (CTD) Niskin bottle rosette was cast to collect a water-column profile, with six 10 L bottles being triggered approximately 2–3 m above the seafloor and three 10 L bottles collected at even intervals throughout the water column (Früh-Green et al., 2017b). Additionally, a 4 L bucket was used to collect surface seawater surrounding the ship. Ten-liter Niskin bottles were also attached to each drill and were used to collect bottom seawater during and immediately after drilling. A total of 109 seawater samples were collected during IODP Exp. 357. All seawater samples were filtered with a peristaltic pump through 0.22 µm Sterivex™ filters onboard the ship and stored at -80 ∘C until DNA extraction (Motamedi et al., 2020).

The drill grease and oil samples were collected in sterile 15 mL tubes directly from their original product packaging. Additionally, one methanol-soaked filter used to wipe down a stainless-steel core liner, plastic shards shaved from a plastic core liner, and samples of spray paint that were used on the drill were collected. These samples were all stored at -80 ∘C until extraction. Two sets of these samples were collected, one for testing of organic chemical contamination (Hickok et al., 2018) and the other for microbial contamination testing, described here. All grease/oil extractions took place in a HEPA-filtered room using an aseptic technique. The lab bench was wiped down with 80 % ethanol prior to each set of extractions. Lab air samples were collected by filtering air through 0.1 µm Puradisc 25 mm PTFE syringe filters (GE Healthcare Whatman, Pittsburgh, Pennsylavania, USA) by a dual-head Air Cadet Model 420-2901-00FK (Thermo Fisher Scientific, Waltham, Massachusetts, USA). The lab air was vacuumed through a total of three different filters for 9 h each and was combined during DNA extraction. DNA extraction and purification were performed using the same protocols and reagents as outlined in Motamedi et al. (2020).

Test DNA extractions were performed to determine the best protocol for use on the industrial grease and oil samples. For the test samples, LMX “Red” Grease (Plews & Edelmann, IL, USA), WD-40 spray (CA, USA), and mineral oil were selected for their accessibility and similarity to industrial grade oils and greases used during IODP Exp. 357. The quality and quantity of extracted DNA from six different protocols were evaluated (Table 1). Two sets of extractions were performed: the first set were unaltered test samples and the second set were spiked with 10 µL of a turbid suspension of E. coli cells to assess DNA recovery. Once the samples were extracted, both sample sets underwent DNA purification via 2x SPRI beads (Rohland et al., 2012). After each step in the purification process, the samples were quantified with a Qubit fluorometer (Invitrogen, CA, detection limit of 50 ng mL-1) and NanoDrop spectrophotometer (Thermo Fisher Scientific, MA) to assess DNA quantity and quality (Table 1). Of the six DNA-extraction methods, the FastDNA^® SPIN Kit (Qbiogene Inc., CA) was the only method to yield DNA from all three test samples.

DNA was isolated from the Exp. 357 grease and oil samples using the FastDNA^® SPIN Kit according to manufacturer recommendations, except that a Mini-Beadbeater-16 (Biospec Products, OK) was used in place of the FastPrep^® instrument. The quantity of starting material for the extraction ranged from 0.25 to 1.0 g (Table 2). Extracted DNA was purified via 2x SPRI beads (Rohland et al., 2012). If replicate extractions of the same sample were possible, the extracted DNA from those replicates was pooled together during the DNA purification step. Additionally, six blank samples of Invitrogen UltraPure™ distilled water were extracted alongside the grease and oil samples as an additional precaution to test for kit contamination, potential contamination introduced during the extraction process, and sequencing contamination (Salter et al., 2014). The extraction of DNA from rock core, seawater, and laboratory air samples was previously described in detail in Motamedi et al. (2020).

Purified DNA preparations from IODP Exp. 357 rock cores, seawater, laboratory air, and greases and oils were sent to the Michigan State University Research and Technology Support Facility Genomics Core for sequencing of the V4 region of the 16S rRNA gene using the duel-indexed Illumina fusion primers 515F-806R (Kozich et al., 2013). All grease and oil samples were submitted for sequencing twice (i.e., sequencing replicates), and the results from both replicates are included in our analysis, with the exception of a single sample, GREd003, one replicate of which was determined to have been compromised during sequencing. Sequences from seawater and rock core samples were previously reported in Motamedi et al. (2020). Analysis of the 16S rRNA gene amplicon sequences from the greases and oils, rock cores, seawater, laboratory air, and extraction blanks was conducted with the mothur (v.1.39.5) software platform (Schloss et al., 2009). Sequences with > 8 homopolymers and > 0 ambiguous bases were removed from downstream analyses, and the sequences were then pre-clustered with the mothur command pre.cluster (diffs = 1) to remove rare sequences most likely created by sequencing errors (Schloss et al., 2011). Operational taxonomic units (OTUs) were formed with a 97 % similarity threshold using the VSEARCH DGC clustering algorithm (Rognes et al., 2016) in mothur. Of the 31 406 783 paired sequences, 75 189 OTUs were identified among the greases and oils, seawater, rock cores, laboratory air, and extraction blanks. Taxonomic classification of all OTUs was performed with mothur using the SILVA reference alignment (SSURefv132) and taxonomy outline (Pruesse et al., 2012). The proportion of contamination from seawater, laboratory air, or industrial grease and oil into each rock core sample was estimated using SourceTracker2 v.2.0.1 (Knights et al., 2011). All sequence data are available via NCBI SRA BioProject PRJNA575221.

To our knowledge, the extraction of DNA from industrial greases and oils has not been previously published, either in the context of scientific drilling projects or in other uses of industrial lubricants. DNA-extraction protocols have been published for testing the integrity of food oils such as olive oil (Busconi, et al., 2003; Consolandi, et al., 2008; Testolin and Lain, 2005) and soybean oil (Pauli et al., 1998), but not industrial oils. Multiple DNA-extraction protocols were evaluated for this study using three test samples (LMX “Red” Grease, WD-40 spray, and mineral oil) that were spiked with E. coli cells before DNA extraction. The MPBio FastDNA^® SPIN Kit (Qbiogene, Inc., CA) was the only method able to extract detectable DNA from all three of the test samples (Table 1). Extractions from these test samples highlighted viscosity as a key challenge for adapting extraction protocols for greases and oils. In general, less viscous oils were easier to extract than the more viscous grease samples due to difficulties in implementing the physical lysing of thick greases. 16S rRNA gene amplicon sequencing of the spiked test samples confirmed that the recovered DNA was dominated by E. coli (data not shown).

To assess the potential of drilling equipment to introduce contamination into drill core samples, we collected 20 samples of greases, oils, plastic shards, spray paint, and a cotton filter that wiped a stainless-steel core liner, all of which were used during IODP Exp. 357 (Table 2). All materials that were sampled had some interaction with the drills and introduced a potential for contamination into the rock cores. Six samples of Invitrogen UltraPure™ distilled water were extracted alongside the grease and oil samples to account for any possible contamination introduced from the DNA-extraction kit or during the extraction and sequencing process.

From these 26 samples, we obtained a total of 4 339 588 paired sequences of 16S rRNA gene amplicons, which were clustered into 5629 OTUs at a 97 % sequence similarity threshold. Any OTUs detected in the DNA-extraction blank samples were removed from the dataset, leaving 4694 OTUs (Table S1 in the Supplement). Gammaproteobacteria constituted the highest percentage of taxa in the oil and grease samples (32 % of total sequences; Fig. 1), and the most abundant Gammaproteobacteria OTUs could not be classified below the class level (Table S1). Alphaproteobacteria (17 % of total sequences) were primarily represented by Sphingomonadaceae, Rhodobacteraceae, and Acetobacteraceae. Bacteroidia (10 % of total sequences) were primarily represented by Flavobacteriaceae, Spirosomaceae, and Hymenobacteraceae. Betaproteobacteriales (4 % of total sequences, but note that Betaproteobacteriales are classified as an order within class Gamamproteobacteria in the SILVA taxonomy) were dominated by various genera of the Burkholderiaceae.

DNA sequences in the oils and greases had high similarity to sequences from a wide range of environments, including soil (e.g., NCBI accessions HM104622, MG716681, AM940870, KP786168, NR_163645), glaciers (HQ333317, MN880348), lake sediments (MT067094), a geothermal plant (KY077452), a shallow marine hydrothermal vent (GU369930), and seawater (JN233022, KX177824). Sequences associated with petroleum-contaminated environments were also identified (EU328045, KY190357). Many of the matching sequences were associated with drilling projects, such as an IODP borehole (KR072759), marine sediments (CP004387, MF977474), a continental borehole (KP901594), and groundwater wells (KC606558). A few OTUs obtained from a swipe of a core liner barrel matched those from continental subsurface studies (MT067098, HM185963, HM641526). Notably, several OTUs that were abundant in the grease that is used on RD2's umbilical cord were nearly identical to clones from a deep-sea drilling and coring contamination study (Yanagawa et al., 2013), notably including those recovered from the drilling fluid in that study (e.g., AB824901). A summary of best sequence matches is provided in the Supplement (Table S2).

All of the most abundant grease/oil OTUs were also identified in samples of seawater collected during the expedition. Because the greases and oils were sampled directly from their commercial product packages (except for the core liner swab and paint chips), contamination from seawater into the grease/oil samples seems unlikely. The grease/oil OTUs that are most abundant in seawater samples were almost exclusively derived from GREd008 (AquaShield, lubrication for MeBo core barrel threads) and GREd015 (RD2 umbilical cord grease). Furthermore, seawater samples collected with Niskin bottles mounted on MeBo and RD2 were especially likely to contain OTUs from the AquaShield grease (Table S1). However, overall, these potential grease/oil contaminants represent a small fraction of the total sequence dataset from seawater (< 1 % of all sequence counts).

The rock core samples collected during IODP Exp. 357 were exposed to potential contamination sources before, during, and after drilling (Fig. 2). The extent of DNA contamination from seawater into the rock cores was investigated by Motamedi et al. (2020), and here, we extend that analysis to include grease and oil samples as additional potential sources of contamination.

Of the 4694 OTUs identified in all grease and oil samples, 565 OTUs were also identified in at least one rock core sample from IODP Exp. 357 (Table S3). However, 86 of these OTUs were also identified in samples of the ambient lab air, suggesting that some of these sequences represent general contamination from dust particles during laboratory handling. In addition, the taxonomic classifications of many of these OTUs suggest that they are derived from commercial reagents (e.g., Burkholderia) or the human microbiota (e.g., Enterobacteriaceae), based on previous studies (Sheik et al., 2018; Salter et al., 2014), even though they were not detected in the extraction blanks or ambient lab air during our study. OTUs that are suspected to be contaminants on the basis of their taxonomic classification are highlighted in (but not removed from) the tables in the Supplement.

The remaining 479 OTUs that represent potential grease/oil contaminants of the rocks (Table S3) comprise 16 % of the total OTUs and 24 % of the total sequence counts in the rock core samples. However, most (90 %) of these sequence counts in rocks are contributed by OTUs that were found in low abundance in our samples of greases and oils (< 100 total counts across all GRE samples), casting doubt that the greases and oils were the source of most of these contaminants into the rock cores. Abundant OTUs from greases and oils were generally very rare in the rock cores.

We estimated the proportion of DNA sequences from each rock core sample that could be attributed to each potential source of contamination (i.e., seawater, laboratory air, or drill grease and oil) using SourceTracker2 (Fig. 3). OTUs with “unknown” sources could not be assigned to a single contamination source and may represent true inhabitants of the rock cores. Lab air was the largest source of contamination into the rock cores, and contamination from other sources was minimal. Greases and oils were estimated to contribute at most a few percent of the sequences in each rock core sample, and their contribution was nearly zero in many of the samples. Nevertheless, the detectable levels of contamination from grease/oil and seawater are notable, considering the extensive precautions employed during handling and processing of the rock core samples (Früh-Green et al., 2017a, 2018; Hickok et al., 2018; Motamedi et al., 2020). These precautions (including the use of bottom seawater as the drilling fluid, immediate freezing of core samples, and shaving of core exteriors with a sterile rock saw in a dedicated facility) are not practical for many drilling projects, suggesting that these contamination levels may be higher in other studies.

We assembled a final list of 27 likely contaminant OTUs from greases and oils (Table S4 and summarized in Table 3) based on their absence in extraction blanks and lab air and their much higher abundance in greases and oils compared to seawater (i.e., > 5 × greater abundance and > 500 total counts in greases and oils; see Table S3 for numbers). These likely contaminants were mostly derived from GREd001 (hydraulic oil used on both RD2 and MeBo) and GREd015 (RD2 umbilical cord grease) and were also moderately abundant in GREd004 (K Nate grease), GREd006 (electrical contact grease), and GREd008 (AquaShield grease). They represent 8 phyla and 10 classes, with Clostridia the most frequently appearing. As noted in Sect. 3.2, the most abundant of these sequences had high similarity to database sequences reported from a wide range of environments.