ABSTRACT
Western Canada produces large amounts of bitumen, a heavy, highly weathered crude oil. Douglas Channel and Hecate Strait on the coast of British Columbia are two water bodies that may be impacted by a proposed pipeline and marine shipping route for diluted bitumen (dilbit). This study investigated the potential of microbial communities from these waters to mitigate the impacts of a potential dilbit spill. Microcosm experiments were set up with water samples representing different seasons, years, sampling stations, and dilbit blends. While the alkane fraction of the tested dilbit blends was almost completely degraded after 28 days, the majority of the polycyclic aromatic hydrocarbons (PAHs) remained. The addition of the dispersant Corexit 9500A most often had either no effect or an enhancing effect on dilbit degradation. Dilbit-degrading microbial communities were highly variable between seasons, years, and stations, with dilbit type having little impact on community trajectories. Potential oil-degrading genera showed a clear succession pattern and were for the most part recruited from the “rare biosphere.” At the community level, dispersant appeared to stimulate an accelerated enrichment of genera typically associated with hydrocarbon degradation, even in dilbit-free controls. This suggests that dispersant-induced growth of hydrocarbon degraders (and not only increased bioavailability of oil-associated hydrocarbons) contributes to the degradation-enhancing effect previously reported for Corexit 9500A.
IMPORTANCE Western Canada hosts large petroleum deposits, which ultimately enter the market in the form of dilbit. Tanker-based shipping represents the primary means to transport dilbit to international markets. With anticipated increases in production to meet global energy needs, the risk of a dilbit spill is expected to increase. This study investigated the potential of microbial communities naturally present in the waters of a potential dilbit shipping lane to mitigate the effects of a spill. Here we show that microbial degradation of dilbit was mostly limited to n-alkanes, while the overall concentration of polycyclic aromatic hydrocarbons, which represent the most toxic fraction of dilbit, decreased only slightly within the time frame of our experiments. We further investigated the effect of the oil dispersant Corexit 9500A on microbial dilbit degradation. Our results highlight the fact that dispersant-associated growth stimulation, and not only increased bioavailability of hydrocarbons and inhibition of specific genera, contributes to the overall effect of dispersant addition.
INTRODUCTION
Western Canada possesses large deposits of oil sands. Petroleum recovered from these oil sands comes in the form of highly weathered, viscous bitumen. To enable its efficient transport in pipelines, bitumen is diluted with low-viscosity gas condensates, creating a crude oil blend termed diluted bitumen (dilbit).
The Douglas Channel is a deep-water fjord system at the northwest coast of Canada that discharges into the open waters of the Hecate Strait (Fig. 1A). The town of Kitimat is located at the head of the Kitimat Arm of Douglas Channel. The port of Kitimat was proposed as a possible marine terminal for the Enbridge Northern Gateway Pipeline, a pipeline that was planned to facilitate the export of dilbit from Canada to Asian markets. These plans would result in heavy oil tanker traffic through the Douglas Channel and the Hecate Strait and hence would increase the risk of dilbit spills into these waters.
Sampling campaigns in the Douglas Channel and the Hecate Strait (BC, Canada) and experimental design of microcosm incubations. (A) The Douglas Channel is indicated in light blue. Sampling sites are marked with filled (black) circles. The location of the town Kitimat is indicated by an open (white) circle. Incubation times shared by setups from all three sampling campaigns are shown in boldface. (B) Overview of the experimental design of microcosm setups. Abbreviations: AWB, Access Western Winter Blend; CLWB, Cold Lake Winter Blend; CLSB, Cold Lake Summer Blend; GC/MS, gas chromatography-mass spectrometry for petroleomics; MoBio, molecular biology to characterize microbial communities. Maps were created in R with data from Natural Earth (https://www.naturalearthdata.com) and GADM (https://gadm.org).
This study investigated the fate of dilbit in Douglas Channel and Hecate Strait waters in order to assess the potential impact of a dilbit spill in this environment. In the event of a spill, microbial communities inhabiting the water surface layer are the first to come into contact with the dilbit and therefore form the first line of defense before the dilbit reaches deeper water layers, the sediment, and coastal beaches and can enter the marine food web. Previous research suggests that spilled, fresh dilbit will likely float in seawater for several days and would thus remain in the surface water layer until it weathers and interacts with suspended particles (1). We therefore focused on the potential of these surface communities to naturally degrade spilled dilbit. Surface water microcosms were prepared from a total of three stations from Douglas Channel and Hecate Strait. The microcosms were amended with three typical western Canadian dilbit blends that would be transported along the Douglas Channel/Hecate Strait shipping route. Seasonal variation of dilbit degradation potential was investigated by comparing microcosms prepared in summer and in winter. Additional microcosms were prepared to investigate the effect of the dispersant Corexit 9500A on the microbial degradation of dilbit. Dilbit degradation was monitored using chemical analyses. Corresponding microbial communities were characterized using 16S rRNA gene amplicon sequencing supplemented with shotgun metagenomic and metatranscriptomic data.
RESULTS AND DISCUSSION
Natural attenuation as a means to mitigate the environmental impacts of a dilbit spill.Microcosm incubations were set up with water samples from stations located in Douglas Channel and Hecate Strait, BC, Canada. The water samples were obtained during three sampling campaigns: two campaigns during summer (in 2014 and 2015) and one during winter (in 2016). A graphical overview of the performed sampling campaigns as well as the design of the microcosm setups is depicted in Fig. 1. Microcosms were amended with three dilbit blends expected to be relevant in case of a dilbit spill in the Douglas Channel and the Hecate Strait: Access Western Winter Blend (AWB), Cold Lake Winter Blend (CLWB), and Cold Lake Summer Blend (CLSB). The tested dilbit blends usually contain 52% to 74% (wt/wt) resins and asphaltenes (2). Resins and asphaltenes are considered highly resistant to degradation (3) and would therefore require physical removal for cleanup. Besides resins and asphaltenes, n-alkanes and polycyclic aromatic hydrocarbons (PAHs) form the remaining fraction of dilbit. These are, in general, susceptible to biodegradation by microorganisms and were hence the focus of this study. Based on chemical analyses of microcosms at T = 0, the hydrocarbon fraction of AWB consisted of 66% ± 4% (mean ± standard deviation) n-alkanes and 31% ± 4% PAHs, the hydrocarbon fraction of CLWB consisted of 56% ± 3% n-alkanes and 43% ± 3% PAHS, and the hydrocarbon fraction of CLSB consisted of 32% ± 1% n-alkanes and 66% ± 1% PAHs (Fig. 2B).
Hydrocarbon loss in microcosms after 28 days. (A) Box plot of the mean loss (n = 3) of n-alkanes, naphthalenes, and other polycyclic aromatic hydrocarbons (PAHs) across biotic and sterile setups after 28 days of incubation at ambient water temperatures. (B) Exemplary hydrocarbon loss over time as observed for 2015 microcosms prepared with water from station FOC. The data shown originate from microcosms without dispersant addition. Bar plots at T = 0 show the typical compositions of the dilbit blends used. Abbreviations: AWB, Access Western Winter Blend; CLWB, Cold Lake Winter Blend; CLSB, Cold Lake Summer Blend.
Dilbit-amended microcosms were prepared using surface water sampled in summer (2014 and 2015) and in winter (2016) and incubated at approximately in situ temperatures (15°C in summer; 7 to 8°C in winter) for up to 42 days. In the majority of summer and winter microcosms, ≥84% of the initial amount of n-alkanes (expressed in ng liter−1/ng liter−1 hopane in order to correct for abiotic dilbit losses) was lost after 28 days (Fig. 2A; see Fig. S1 in the supplemental material). Alkane loss in the biotic microcosms was significantly higher than that in sterile controls (Fig. S1), so the observed loss can be attributed to microbial degradation. As expected for microbial alkane degradation, the C17/pristane and C18/phytane ratios measured for the biotic microcosms generally decreased over time (not shown). Estimated times required for the concentration to decline to half of the initial value (DT50 [half-life] values) for alkane degradation were between 4 and 35 days and varied between seasons, years, stations, and the type of added dilbit blend (see Data Set S1 in the supplemental material). n-Alkane degradation in the Douglas Channel/Hecate Strait summer microcosms (incubated at 15°C) appeared to be on average faster (mean DT50 of 8.7 days) than dilbit-associated alkane degradation in 19°C flume tanks inoculated with seawater from Bedford Basin, NS, Canada (mean DT50 values of 20.6 to 26.3 days [4]).
Microbial degradation of PAHs depends highly on their water solubility, which is a function of the number of aromatic rings as well as the number of (hydrophobic) alkyl groups. In general, PAHs with more than three rings are very poorly soluble in water (5) and hence are less available for microbial degradation. Confirming this, we observed significant loss of only two-ring (naphthalenes) and three-ring (fluorenes, dibenzothiophenes, and phenanthrenes) PAHs. More complex PAHs such as four-ring chrysenes, pyrenes, and naphthobenzothiophenes showed no significant loss within the time frame of our incubations (Fig. S1). A comparison between rates of loss of two-ring and three-ring PAHs between biotic microcosms and sterile controls indicated that PAH loss in biotic microcosms can be attributed to microbial degradation.
Microbial degradation of PAHs expectedly varied by compound. Of all PAHs, naphthalene and its methylated derivatives most consistently showed a significantly increased degradation compared to sterile controls. Overall, ≥25% of the initial amount of naphthalenes was lost after 28 days (Fig. 2A). The degradation of naphthalenes appeared to be a function of methylation, where decreasing degradation rates were associated with increasing numbers of methyl groups (see Fig. S2 in the supplemental material). Moderate degradation of the PAHs fluorene, dibenzothiophene, and phenanthrene as well as their methylated derivatives was also consistently observed. As was observed for naphthalenes, degradation of these compounds also appeared to be a function of methylation (see Fig. S3 to S5 in the supplemental material): the degradation of fluorenes, dibenzothiophene, and phenanthrene with more than one methyl group was in general not significantly different between biotic microcosms and sterile controls. Similar decreasing degradation rates with increasing numbers of methyl groups have been reported in previous studies focusing on the microbial degradation of other types of crude oil (6–8).
Due to the degradation of mainly n-alkanes and only selected PAHs, the concentration of hydrocarbons remaining in the microcosms strongly depended on the initial dilbit composition. In summary, after 28 days of incubation, the concentration of residual hydrocarbons was reduced to 24% to 47% (based on the starting amount) in microcosms supplemented with AWB, to 33% to 54% with CLWB, and to 58% to 93% with CLSB (Fig. 2B).
Potential evolution of dilbit toxicity.Polycyclic aromatic hydrocarbons represent the most toxic fraction of crude oil (9). Degradation-resistant PAHs, i.e., PAHs that either were too methylated or featured more than three rings, made up the majority of the PAH fraction of the tested dilbit blends (80% for AWB, 67% for CLWB, and 75% for CLSB). As a result, the overall amount of PAHs changed only slightly over the microcosm incubation period. It is currently difficult to predict how this affected the overall toxicity of dilbit during the course of our incubations. Toxicity of crude oil to fish has been mostly linked to 3- to 5-ring PAHs, with methylated and nonmethylated PAHs showing different types of toxicity (10, 11). As most of the 3- to 5-ring PAHs were not significantly degraded in this study, this suggests that dilbit toxicity to fish might not have changed significantly.
Toxicity of PAHs to microorganisms and microalgae appears to be mainly a function of water solubility, with less-water-soluble PAHs being more toxic (12). Because water solubility decreases as a function of methylation and methylated PAHs were mostly not degraded, we expect that dilbit toxicity to microorganisms and microalgae may also not have changed throughout our incubations. Ultimately, future studies specifically targeting the evolution of dilbit toxicity throughout the biodegradation progress will be necessary to determine how effective this (biodegradation) process is to mitigate the environmental toxicity of a dilbit spill.
Cell growth in microcosms is stimulated by dispersant alone and by dilbit addition.The yield of DNA recovered from microcosms and seawater samples was used to estimate corresponding cell densities and thereby determine bacterial growth. In general, increases in estimated cell densities in early-stage microcosms (after 3 or 5 days of incubation) appeared to be mostly associated with the addition of the dispersant Corexit 9500A, while increases in estimated cell densities at later stages (after 28 or 42 days) appeared to be mainly associated with dilbit addition (Fig. 3; see Data Set S3 in the supplemental material). In early-stage microcosms, 5 out of 6 microcosm setups amended with dilbit and dispersant, and 4 out of 6 with dispersant only, showed a significant increase compared to the starting cell density. In comparison, only 2 out of 6 microcosm setups amended with dilbit only, and 3 out of 6 setups with neither dilbit nor dispersant, showed a significant increase in estimated cell density at this incubation stage. Estimated cell densities of all dilbit-amended microcosms (irrespective of concurrent dispersant addition) had significantly increased after 28 or 42 days of incubation. Two dilbit-free control setups (seawater amended with mineral nutrients) showed a significant increase in estimated cell densities after 28 or 42 days. The three main ingredients of Corexit 9500A are light petroleum distillates (50%), glycols (40%), and dioctyl sulfosuccinate (DOSS) (10%) (13). All of these ingredients have previously been shown to serve as substrates for microbial growth (14–16). The stimulation of cell growth by Corexit 9500A is therefore not surprising, and similar findings have been reported elsewhere (17–19).
Development of extractable DNA and enrichment of potential hydrocarbon degraders. (A) Development of the amount of extractable DNA per milliliter of seawater (T = 0) or microcosm sample. Corresponding cell densities were estimated based on the amount of extractable DNA. (B) Sum of relative abundances of the potential hydrocarbon-degrading genera Polaribacter, Pseudomonas, Pseudoalteromonas, Marinobacter, Oleispira, Colwellia, Cycloclasticus, Thalassolituus, Glaciecola, Alteromonas, Alcanivorax, Oceaniserpentilla, Oleibacter, and Oleiphilus. Treatments are color coded. Panel columns represent sampling times for microcosms, i.e., T = 0, after 3 or 5 days (early stage), and after 28 or 42 days (late stage). Panel rows represent sampling years and stations.
Development of richness and diversity in microcosms.The species richness and diversity of microcosms both decreased significantly over time (see Fig. S6 and Data Sets S4 and S5 in the supplemental material). The richness decreased from a median of 891 species at T = 0 to a median of 274 species in the late-stage microcosms. At the same time, the effective number of species decreased from a median of 27 to a median of 8. Exceptions to this trend were microcosms set up in 2015 from station HEC. Here, species richness and diversity were already comparably low in the source seawater (richness, 551; effective number of species, 9) and did not change significantly over time. Inspection of rarefaction curves (data not shown) of 2015 microcosms from station HEC suggested that the corresponding 16S rRNA gene amplicon libraries (which formed the basis of richness and diversity estimations) were sufficiently sampled. Although not consistent for all microcosms, dispersant addition appeared to cause a faster initial loss of richness and diversity. Differences between richness and diversity of dilbit-amended microcosms with and without dispersant were not statistically significant after 28 or 42 days (Data Sets S4 and S5).
Hierarchy of factors driving microbial community composition of dilbit-amended microcosms.The identification of community-shaping factors was challenging; no clear patterns became apparent when analyzing the full data set using multivariate ordination approaches. We were finally able to identify community-shaping factors by using a hierarchical clustering approach and focusing only on microcosms amended with dilbit (Fig. 4). At the highest level (Fig. 4A), microbial communities formed two clusters: one representing samples from summer microcosms (2014 and 2015) and one representing samples from winter microcosms (2016). Within the summer cluster, microbial communities clustered according to sampling station (Fig. 4B), i.e., separate clusters for the Douglas Channel station FOC and the Hecate Strait station HEC. No clear clustering corresponding to sampling station was apparent for winter microcosms, possibly due to the overall shorter geographic distance between the winter sampling stations FOC and KSK and to both stations being located in the Douglas Channel. Within the three main clusters, incubation time appeared to have been the most important factor shaping the microbial communities (Fig. 4C). In general, microbial communities became less and less similar to their starting communities over time. Within clusters representing the different incubation times, samples appeared to cluster according to dispersant addition (Fig. 4D). Overall, the type of added dilbit (i.e., AWB, CLWB, or CLSB) appeared to have had the least influence on shaping the microbial communities of the microcosms. The observed hierarchy of community shaping factors was supported by permutational multivariate analysis of variance (PERMANOVA) (Table 1).
Grouping of microcosm samples based on neighbor-joining analysis of the proportionality metric ϕ between microbial communities. Each tip (marked with dots) represents a single microcosm sample. Tips are color coded according to sampling year (A), sampling station (B), incubation time (C), or dispersant addition (D).
Influence of factors on the overall microbial communities of dilbit-amended microcosms
Enrichment and succession of potential hydrocarbon degraders.The relative abundance of genera previously reported to be associated with the degradation of hydrocarbons increased with time in dilbit-amended microcosms (Fig. 3B). During the early incubation stages, the relative abundance of these potential hydrocarbon degraders was on average higher in microcosms amended with dilbit and dispersant than in those amended with dilbit only. In the later incubation stages, there was no significantly increased enrichment of potential hydrocarbon degraders in dilbit microcosms with dispersant compared to dilbit microcosms without dispersant. Interestingly, microcosms amended with only dispersant showed an enrichment of potential hydrocarbon degraders similar to that in microcosms amended with both dilbit and dispersant (Fig. 3B).
Similar to the case for whole microbial communities, the composition of the potential hydrocarbon degrader community varied by season, sampling year, sampling station, and incubation period (see Fig. S7 in the supplemental material). The genera Glaciecola, Pseudoalteromonas, Oceaniserpentilla, Polaribacter, and Thalassolituus (only in 2014) were frequently detected in significant proportions in early-stage summer microcosms. In late-stage summer microcosms, potential hydrocarbon-degrading communities were mostly dominated by the genera Alcanivorax, Marinobacter, and in some cases Pseudoalteromonas and Glaciecola. As n-alkanes have been almost completely degraded in late-stage microcosms (Fig. 2), we hypothesize that this community shift represents a shift from a mainly alkane-fueled community to a community fueled mainly by degradation of the more recalcitrant PAHs. In support of this hypothesis, both the genera Marinobacter and Pseudoalteromonas have been reported to harbor PAH-degrading phenotypes (see, e.g., references 20 and 21). The genus Glaciecola has been previously detected (22) in oil-contaminated samples but has so far not been conclusively associated with actual hydrocarbon degradation. Instead, it has been reported to be PAH tolerant (23). Surprisingly, the widespread keystone PAH degradation genus Cycloclasticus (24) was largely absent in late-stage summer microcosms.
Communities of potential hydrocarbon degraders of winter microcosms were fundamentally different from those of summer microcosms. Possibly due to overall lower degradation rates, differences between early- and late-stage microcosms were less pronounced in winter microcosms than in summer microcosms. The dominant genera at both stages were Oleispira, Colwellia, and Pseudoalteromonas. All three genera have been previously reported to contain representatives that perform hydrocarbon degradation at low temperatures (see, e.g., references 25, 26, and 27), which could explain their dominance in dilbit-amended winter microcosms. In addition, we also detected significant proportions of the typically PAH-degrading genus Cycloclasticus in some late-stage winter microcosms.
Degradation and community differences between seasons.n-Alkanes appeared to generally degrade faster in summer microcosms than in winter microcosms, as would be expected due to the higher incubation temperature of the former (Data Set S1). However, except for a single case (microcosms amended with CLSB), actual alkane DT50 (half-life) values in the different seasons were surprisingly similar: 6.4 to 11.8 days for summer microcosms and 8.5 to 11.5 days for winter microcosms. In contrast to n-alkanes, naphthalenes appeared to degrade significantly faster in winter microcosms than in summer microcosms. The corresponding DT25 values ranged from 17.6 to more than 28 days for summer microcosms and from 6.6 to 21.9 days for winter microcosms.
Alkane degradation patterns differed significantly between summer and winter microcosms. Low- and high-molecular-weight n-alkanes (C10 to C35) all appeared to degrade at similar rates in summer microcosms. In contrast, alkane degradation rates in winter microcosms appeared to be a function of alkane chain length, with high-molecular-weight n-alkanes degrading more slowly than those with lower molecular weight (see Fig. S8 in the supplemental material). A similar preferential degradation of low-molecular-weight n-alkanes has been reported for oil-amended microcosms from the Gulf of Mexico incubated at 8°C (17).
Overall microbial communities as well as communities of potential dilbit degraders were fundamentally different in summer and winter microcosms (Fig. 4 and S7). A focused analysis of the microbial communities at T = 0 (data not shown) yielded a clustering similar to that found when using the complete community data set of all dilbit-amended microcosms (Fig. 4). This implies that community differences were determined not only by the difference in incubation temperature but also by seasonally different starting communities.
Effect of dispersant addition on dilbit degradation and community development.The application of dispersants is one means to try to mitigate the environmental effects of marine oil spills. Dispersants are intended to break up oil slicks into smaller droplets that are then more available for oil-degrading microorganisms. Corexit 9500A was the main dispersant used to mitigate the effects of the Deepwater Horizon oil spill (28). The ability of the dispersant Corexit 9500A to enhance biodegradation of crude oil has been the topic of several recent studies (8, 17, 25, 29–31). The present study investigated the effect of Corexit 9500A addition on the biodegradation of dilbit blends.
Microcosms with and without added Corexit 9500A were compared to determine the effect of this dispersant on the degradation of n-alkanes and naphthalenes. This resulted in a total of 16 comparisons: one for each unique sampling year, sampling station, and dilbit-type combination. Overall, the type of effect—i.e., accelerating, neutral, or inhibiting—followed no discernible pattern. We hypothesize that this lack of a predictable pattern is mainly a result of the considerable dispersion of the petroleomic data even between replicates, which in turn was caused by a more or less stochastic assembly of the underlying hydrocarbon degrader communities (see “Microcosm variability and implications for predictability of in situ biodegradation of dilbit” below) and other confounding factors. Despite the dispersion of the data, it appeared that the addition of Corexit 9500A mostly had either no effect or increased the degradation of n-alkanes and naphthalenes. Dispersant addition significantly decreased the DT50 values of n-alkanes in 6 out of 14 comparisons. In another 6 comparisons, the dispersant had no significant effect on alkane degradation, and in only 2 comparisons was the DT50 value significantly increased by dispersant addition. The biggest effect of dispersant addition, a DT50 decrease by 23.5 days, was observed in winter microcosms with water from station FOC and supplemented with CLSB.
The effect of dispersant addition on the degradation of naphthalenes was similarly variable as for n-alkanes (see Data Set S2 in the supplemental material). In 8 of 14 comparisons, dispersant addition had no significant effect. In 3 comparisons, dispersant addition decreased the degradation rate of naphthalenes, while in another 3 comparisons, the degradation rate of naphthalenes was increased by dispersant addition. All three cases of enhanced degradation of naphthalenes were observed for winter microcosms. In these microcosms, dispersant addition also enhanced the degradation of the three-ring PAHs fluorene, phenanthrene, and dibenzothiophene as well as of some of their methylated derivatives (Fig. S3 to S5). This trend was most evident for microcosms amended with the CLWB dilbit blend.
The trajectories of microbial community development for a given year and sampling station indicated that dispersant addition rather than dilbit addition was the most important factor shaping microbial communities during early incubation stages. This was especially obvious for 2014 summer microcosms from station FOC and for winter microcosms; in these microcosms, there was no significant difference between the early-stage communities of microcosms amended with dilbit and dispersant and those amended with dispersant alone (Fig. 5; see Fig. S9 in the supplemental material). Similarly, microbial communities from microcosms amended with dilbit alone appeared to be more similar to those from microcosms without both dilbit and dispersant than to those amended with dilbit and dispersant (Fig. 5 and S9). These observations are supported by PERMANOVA analysis (see Data Set S6 in the supplemental material), which indicates that dispersant addition is a significant factor influencing the dissimilarity between samples and generally accounts for more of the variation between samples at this early stage than dilbit addition (Data Set S6).
Effect of dispersant addition on microbial communities using the example of 2016 winter microcosms from station FOC. (A) Nonmetric multidimensional scaling (nMDS) ordination of microbial communities of 2016 winter microcosms from station FOC. Treatments are color coded. Community dissimilarities are based on the proportionality metric ϕ. Ellipses indicate 95% confidence range. (B) Differentially abundant genera in early-stage (after 5 days of incubation) 2016 winter FOC microcosms. Nondifferentially abundant genera are not shown. Plus and minus signs in the color-coding legend indicate increased abundance in treatments with (+) and without (−) dispersant, respectively. (C) Abundances of genes potentially involved in hydrocarbon degradation in early-stage 2016 winter FOC microcosms. Heat map tiles are colored based on counts-per-million (cpm) values as determined by metagenomic analysis. For visualization purposes, cpm values were centered and scaled across rows, i.e., on a per-gene basis. Statistically significantly enriched gene abundances were determined using a false-discovery rate of 0.05. Abbreviations: w/o, without added dilbit; A, Access Western Winter Blend; C, Cold Lake Winter Blend; S, Cold Lake Summer Blend.
In general, this observation could have been caused by either (i) dispersant-induced inhibition or (ii) dispersant-induced stimulation of the same community members in dilbit-plus-dispersant and dispersant-only microcosms. We sequenced and analyzed metagenomes and metatranscriptomes of microcosms to further investigate this observation. The single-copy rpoB gene (encoding the β subunit of bacterial RNA polymerase) universally present in microorganisms was used to determine the composition (metagenome) and the live fraction (metatranscriptome) of microcosm communities. In general, community dissimilarities based on the metagenomic rpoB data matched very well with those of the corresponding 16S rRNA gene amplicon data (Fig. 5; see Fig. S9 and S10 in the supplemental material). Metatranscriptomic rpoB data showed that the active fraction of the microbial communities from microcosms amended with dilbit and dispersant were most similar to those from microcosms amended with dispersant only. At the same time, the active fractions of the microbial communities from microcosms amended with only dilbit were most similar to those without both dilbit and dispersant. A PERMANOVA analysis supports these results (see Data Set S7 in the supplemental material).
Differential abundance analysis between dispersant-amended (both with and without dilbit) and dispersant-free early-stage microcosms revealed several genera that were frequently more abundant in the dispersant-amended microcosms (Fig. 6; see Data Sets S10 to S12 in the supplemental material). An analysis of the operational taxonomic units (OTUs) making up these genera confirmed a selective enrichment of the same OTUs in dispersant-amended microcosms with and without added dilbit (Fig. S7). Interestingly, most of the dispersant-associated genera are usually also associated with hydrocarbon degradation, namely, Colwellia, Oceaniserpentilla, Oleispira, Shewanella, Oleibacter, Thalassolituus, Flavobacterium, and Pseudoalteromonas. Metagenomic analysis revealed a similar enrichment pattern of hydrocarbon degradation genes between microcosms with dilbit and dispersant and microcosms with dispersant alone (Fig. 5C; see Fig. S11 to S13 in the supplemental material).
Dispersant-associated genera after 3 or 5 days of incubation. Shown are genera that were differentially abundant in dispersant-amended (with and without dilbit) microcosms compared to dispersant-free microcosms. Only genera that were repeatedly detected as differentially abundant are shown.
A dispersant-stimulated enrichment of potential hydrocarbon degraders, such as Colwellia and Rhodobacteraceae, has also been observed in other studies (14, 18, 25, 32). This observation is not surprising, as Corexit 9500A contains a large hydrocarbon fraction in the form of alkane side chains of DOSS and petroleum distillates. Interestingly, however, cell growth and an enrichment of potential hydrocarbon degraders were delayed in microcosms with dilbit but without Corexit 9500A compared to microcosms amended with dilbit and Corexit 9500A. This implies that Corexit-associated hydrocarbons are more bioavailable than dilbit-associated ones. The accelerated enrichment of hydrocarbon degraders in dispersant-amended microcosms did not consistently result in enhanced hydrocarbon degradation in this study. This observation, however, raises the question how much of the previously reported Corexit-associated enhancement of oil degradation (8, 25, 30, 31) can be attributed to an enrichment of hydrocarbon degraders caused by the hydrocarbons contained in Corexit 9500A.
In the later incubation stages, microbial communities from the four types of treatment (i.e., unamended, dilbit only, dilbit plus dispersant, and dispersant only) appeared to be more clearly separated. This holds true for the microbial composition as determined by 16S rRNA gene amplicon data and metagenomic rpoB data, as well as for the active fraction of these communities as determined by metatranscriptomic rpoB data. PERMANOVA analyses indicate that both dilbit and dispersant account for similar fractions of the dissimilarity of microbial communities at this stage (Data Set S7).
Microcosm variability and implications for predictability of in situ biodegradation of dilbit.In this study, microcosms were used to assess the potential for natural attenuation of a marine dilbit spill. Dilbit-degrading microbial communities in the prepared microcosms featured, for the most part, diverging trajectories between replicates: the percentage of OTUs shared between replicates decreased from a median of 93% at T = 0 to 59% after 42 days of incubation (Fig. 7). We hypothesize that this variability was caused by the small population sizes of potential hydrocarbon degraders in the microcosm inocula; potential hydrocarbon-degrading genera that dominated early- and late-stage microcosms were, with few exceptions (95% quantile), either not detectable or very rare (<0.2% relative abundance) in samples from T = 0 (Fig. S7). The recruitment of hydrocarbon degraders from the so-called “rare biosphere” has also been reported in other studies focusing on oil degradation in marine settings (23, 32).
Box plots of similarities between replicates with time. Box plots represent the distributions of OTU percentage shared by all replicates for a given treatment and time point. For the majority of cases, three replicates were used for calculating the metric shown.
Small microbial populations follow a stochastic growth model and have an increased extinction risk (33). In the context of our study, we hypothesize that these small-population phenomena introduced a high degree of stochasticity in terms of which of the initially small oil degrader populations will eventually dominate a given microcosm. The result was microbial communities that were shaped by a more or less stochastic assembly of these initially rare hydrocarbon degraders. As the community assembly process was limited by what was initially present in the microcosm inocula, the resulting communities were also influenced by the spatial and temporal factors that shaped these initial inocula. This is demonstrated by the fact that we were able to identify a hierarchy of environmental factors that shaped microcosm communities (Fig. 4), despite the considerable variation between replicates.
Data from the quantification of hydrocarbons also varied considerably between replicates (see Fig. S14 and S15 in the supplemental material). These data, however, also featured other sources of error (as indicated by differences of replicates at T = 0) as a confounding factor, so that it is currently impossible to determine how much of this variability can be contributed to different underlying degrader communities. The experimental design used, where chemical data and microbial community data originate from separate microcosm flasks, further makes it impossible to directly link degradation rates to specific communities.
Considering the role that stochasticity seemingly played in shaping the hydrocarbon-degrading communities in the microcosms, it is currently difficult to determine how representative the observed degradation rates and community trajectories are of those that would be encountered in situ. Under in situ conditions, the marine water body would function as a continuous seed bank of potential hydrocarbon degraders. We hypothesize that this would largely prevent the occurrence of small-population phenomena. Instead, we expect the in situ community assembly to be determined mainly by how efficiently the different populations of degraders can utilize the available substrates, possibly resulting in higher hydrocarbon degradation rates. Ultimately, future studies that better approximate in situ conditions are necessary to corroborate the results of this study. Results from a study of nutrient-cycling microcosms (34) suggest that an increased microcosm size (and thereby increased overall sizes of hydrocarbon degrader starter populations) would yield more reproducible community trajectories and would be a first step in this direction.
Conclusions.This study showed that indigenous microbial communities of the Douglas Channel and the Hecate Strait have the ability to degrade the hydrocarbon fractions of different dilbit blends. Within the time frame of the present study, this degradation, however, appears to have been limited to n-alkanes and only selected PAHs. The majority of the dilbit PAH fraction remained unchanged after 28 days of incubation. Dilbit-degrading microbial communities of microcosms varied significantly between season, year, and sampling station and even between replicates. Most of the enriched potential hydrocarbon degraders were recruited from the so-called “rare biosphere.” The resulting initially low abundance of potential dilbit degraders most likely caused the observed variability between replicates via small-population-size effects. For the most part, addition of the dispersant Corexit 9500A had an enhancing effect or no discernible effect on the degradation of dilbit hydrocarbons. The greatest effect of dispersant addition was observed for winter microcosms, where it especially enhanced the degradation of PAHs. Microcosms with added Corexit 9500A featured an accelerated enrichment of potential hydrocarbon-degrading genera, which was most likely caused by the degradation of hydrocarbons contained in Corexit. We hypothesize that this priming effect of Corexit 9500A (leading to early high abundances of potential hydrocarbon degraders), and not only increased bioavailability of hydrocarbons, contributes to the previously reported enhancement of oil degradation rates associated with dispersant addition. Future studies will need to establish the importance and magnitude of this effect in oil spill scenarios.
MATERIALS AND METHODS
Field sampling.Surface seawater samples (3- to 5-m depth) were collected using Sea-Bird carousel water samplers (Sea-Bird Scientific, Bellevue, WA, USA) equipped with Niskin bottles. Summer samples were collected in July of 2014 and 2015 from the CCGS John P. Tully at station FOC (53.736°N, 129.030°W) in the Douglas Channel and at station HEC (52.821°N, 129.846°W) in the Hecate Strait near the entrance of Douglas Channel (Fig. 1A). Winter samples from station FOC were collected in March 2016 from the CCGS W.E. Ricker (Fig. 1A). Due to rough weather, it was not possible to obtain samples from the offshore station HEC during the winter 2016 sampling campaign. Station HEC was hence replaced by station KSK (53.480°N, 129.209°W) during this campaign. The stations FOC, HEC, and KSK of this study are identical to the stations FOC1, HEC1, and KSK1, respectively, described previously (35). For a detailed overview of the sampling campaigns, see Fig. 1A. Subsamples for molecular biology analyses of T = 0 were collected in biological triplicates (each representing a separate Niskin bottle) by filtration of 2 liters of seawater onto polyethersulfone membranes (pore size, 0.22 μm; Millipore) immediately following sample collection. Filter membranes were flash frozen using liquid nitrogen and stored at −80°C until processed for nucleic acid extraction.
Preparation of microcosms and general experimental design.A graphical overview of the experimental design of the microcosm setups as well as the campaign-associated incubation times is depicted in Fig. 1. Microcosm experiments were initiated onboard ship as soon as samples were collected. For each sampling event, the contents of multiple Niskin bottle casts were combined to fill an acid-cleaned 20-liter jerrican. This integrated seawater sample was divided into three jerricans that were used for preparing the individual flask microcosms. One hundred milliliters from these blended seawater replicates was transferred to 150-ml baffled flasks. Microcosms were supplemented with 2 ml 1× Bushnell-Haas mineral nutrient broth (Difco, Becton, Dickinson and Company, Mississauga, ON, Canada) to provide sufficient nutrients (N and P) for biodegradation. For each station, triplicate microcosm setups of (i) dilbit only, (ii) dilbit with the dispersant Corexit 9500A (Nalco Energy Services, Burlington, ON, Canada), (iii) Corexit 9500A only, and (iv) seawater only were prepared (Fig. 1B). Microcosm replicates were prepared with water samples from separate jerricans (see above). Microcosms were generally set up with the dilbit blends Access Western Winter Blend (AWB) and Cold Lake Winter Blend (CLWB). Additionally, the Cold Lake Summer Blend (CLSB) dilbit was added to our setup in 2015 and was subsequently used to prepare 2015 summer and 2016 winter microcosms from station FOC. Dilbit blends were provided by the Canadian Department of Fisheries and Oceans through the World Class Tanker Safety System (WCTSS) program. Dilbit blends were artificially weathered prior to use by purging (24 to 48 h) with a gentle stream of nitrogen. The change in mass was recorded to quantify the extent of weathering and amounted to 7% to 10%. Dilbit required heating to 40°C to dispense and was added to a final concentration of 150 ppm. The dilbit-dispersant mixture was prepared at a dispersant-to-oil ratio of 1:20. To determine abiotic dilbit loss, sterile controls for “dilbit only” and “dilbit with Corexit 9500A” microcosms were set up concurrently using the same experimental conditions as for biotic microcosms. Sterile controls consisted of sterile-filtered (pore size, 0.1 μm) seawater amended with sterile (autoclaved twice for 30 min at 121°C) dilbit or dilbit-plus-dispersant mixtures. Autoclaving did not appear to have significantly changed the composition of dilbit (see Fig. S16 in the supplemental material). All microcosm bottles were hermetically sealed after preparation. Microcosms were incubated at the approximate temperature of surface seawater at the time of collection: 15°C for summer microcosms and 7 to 8°C for winter microcosms. Microcosms were continually mixed at 150 rpm using orbital shaker tables. Two parallel sets of microcosms were prepared for (i) molecular biology analyses and (ii) chemical analyses of hydrocarbons (petroleomics).
Samples at all time points were obtained via sacrificial (i.e., destructive) sampling. In general, petroleomic data of microcosms were obtained for 5 time points. Petroleomic data of the initial (2014) microcosm incubations were obtained at 12- to 14-day intervals (for details, see below and Fig. 1A) and for a total of 42 days of incubation. Analysis of these data showed almost complete depletion of n-alkanes after 15 days of incubation. Hence, in order to better capture the early n-alkane degradation kinetics, an additional (early) sampling point after 7 days of incubation was added for the subsequent 2015 microcosm incubations. Due to an expected slower degradation, sampling times for the 2016 winter microcosms were initially planned with 12- to 14-day sampling intervals to mirror those for the 2014 summer campaign. However, respirometric data of the 2016 microcosms showed very little oxygen consumption after 3 days of incubation (the planned first sampling point) (data not shown) and prompted a postponement of the initial sampling time to after 5 days of incubation. Logistics limited us to a total of 3 sampling time points for molecular biology analyses. Sampling times were chosen to capture the microcosm starting communities (T = 0), as well as typical early- and late-stage microcosm communities.
Chemical analyses of microcosms.Hydrocarbon degradation was monitored by sacrificial sampling at 5 time points: 2014, T = 0, 3, 15, 28, and 42 days; 2015, T = 0, 3, 7, 15, and 28 days; 2016, T = 0, 5, 15, 28, and 42 days (see also Fig. 1A). Generally, each time point was represented by three replicates. Microcosms intended for chemical analysis were sacrificed by adding 10 ml of dichloromethane (distilled in glass; Caledon, Georgetown, ON, Canada), followed by vigorous mixing for 30 s, and were finally stored at 4°C until processed for analysis.
Water samples were processed using liquid-liquid extraction (modified version of U.S. Environmental Protection Agency method 3510C). Further details can be found in reference 36. Purified and concentrated extracts were analyzed using high-resolution gas chromatography (GC) (6890 GC; Agilent, Wilmington, DE, USA) coupled to an Agilent 5975B mass selective detector operated in the selective ion-monitoring mode using the following GC (SLB-5ms column, 30 m by 0.25 mm [inner diameter], 0.25-μm film thickness; Supelco, Mississauga, ON) conditions: cool on-column injection with oven track mode (tracks 3°C higher than the oven temperature program), 85°C hold for 2 min, ramp at 4°C min−1 to 280°C, and hold for 20 min. Quantification criteria for PAH were as described previously (29). All petroleomic data were normalized to the conservative biomarker 17α(H),21β(H)-hopane (37, 38) to correct for potential abiotic losses of dilbit and thereby focus only on dilbit biodegradation.
Rates of loss of n-alkanes in different setups were compared using DT50 values, i.e., the time required for the concentration to decline to half of the initial value. The naphthalenes loss rates were compared based on DT25 values due to generally slower loss of these compounds. In order to best represent the variability in the chemical data (see Fig. S14 and S15 in the supplemental material), DT50 values of n-alkanes were analyzed as follows (see also Fig. S17 in the supplemental material). An initial DT50 value was estimated by linear interpolation between mean values of hydrocarbon concentrations (measured in ng liter−1/ng liter−1 hopane) of triplicates. This initial value was used to identify values at actually measured time points left (T = i) and right (T = i + 1) of the estimated DT50 value. All possible combinations of concentrations at these time points were used to calculate a spectrum of loss rates. In cases of triplicate measurements at each time point, this led to a range of 9 loss rates. Rates of cases where the concentration at T = i was lower than that at T = i + 1 (i.e., representing an apparent increase of hydrocarbon concentration over time) were removed from the analysis. Based on the calculated loss rates and three possible starting concentrations (triplicate measurements at T = 0), a range of 27 DT50 values was calculated for each setup. Differences between setups were quantified by subtraction of their corresponding mean DT50 values. The statistical significance of differences between means was determined using the two-sample Wilcoxon rank sum test, as the DT50 values of setups were often not normally distributed (P value of Shapiro-Wilk normality test, ≥0.05). A P value of ≤0.05 was used as the cutoff for statistical significance. Differences between setups where one setup featured a mean DT50 later than the total incubation time, i.e., an extrapolated DT50 value, were always considered significant. Cases where both mean DT50 values were above the incubation time were considered not testable. The DT25 values of naphthalenes were calculated accordingly. All data manipulations as well as statistical analyses were performed using R and the R “stats” package (39).
Characterization of microbial communities.Microbial communities of microcosms were characterized by sacrificial sampling at 3 time points: 2014, T = 0, 3, and 42 days; 2015, T = 0, 3, and 28 days; 2016, T = 0, 5, and 42 days (see also Fig. 1A). Microcosms sampled after 3 or 5 days are referred to as “early-stage” microcosms, and those sampled after 28 or 42 days are referred to as “late-stage” microcosms. In general, each time point was represented by three replicates. Microcosms intended for molecular biology analyses were sacrificed as described for T = 0 samples by filtration onto polyethersulfone membranes (pore size, 0.22 μm; Millipore). Total nucleic acids were extracted from samples as described in detail previously (29). Previous studies have demonstrated a strong correlation between sample cell densities and the corresponding yield of extractable DNA (40, 41). Based on this, cell densities of microcosms were estimated from the yields of extracted DNA and assuming an average DNA content of 4 × 10−15 g of DNA per cell (derived from an assumed average genome size of 4 Mbp and a ratio of ≈1 fg of DNA per 1 Mbp; the ratio of 1 fg of DNA per 1 Mbp is based on the average molecular weight of a base pair of DNA divided by the Avogadro constant, i.e., 618 g mol−1/6.022 × 1023 mol−1 ≈ 1 × 10−21 g per bp). The statistical significance of differences between estimated cell densities was determined using a one-tailed, two-sample Wilcoxon rank sum test.
Basic characterization of microbial communities was carried out by 16S rRNA gene sequencing after PCR amplification using the primer set 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′); the primer set targets both Archaea and Bacteria (42). PCR amplification was performed using the HotStarTaq Master Start polymerase (Qiagen, Toronto, ON, Canada) as follows. Reactions were performed in a 25-μl volume containing 5 to 25 ng of template DNA, 0.2 μM each primer, 0.5 mg/ml bovine serum albumin, and 1.24 μM (each) antimitochondrial (5′-GGCAAGTGTTCTTCGGA-3′) and antiplastid (5′-GGCTCAACCCTGGACAG-3′) peptide nucleic acid PCR clamps (43). Amplification was performed using an initial denaturation for 15 min at 95°C followed by 25 cycles of 45 s at 95°C, 10 s at 78°C, 60 s at 55°C, and 90 s at 72°C and a final elongation for 7 min at 72°C. Success of the PCR amplification was evaluated by gel electrophoresis. PCR amplicons were purified using the Macherey-Nagel NucleoMag NGS Clean-Up and Size Select kit (d-MARK Biosciences, Toronto, ON, Canada). PCR products were indexed, and equal amounts of each indexed PCR product were pooled and finally sequenced using the Illumina MiSeq platform and the 500-cycle MiSeq reagent kit v2.
Reads were quality checked, paired-end assembled, and clustered at 97% sequence identity using the rRNA short amplicon analysis pipeline of the National Research Council (NRC) (Montreal, Canada) as described in detail previously (44, 45). The resulting operational taxonomic units (OTUs) were assigned to a taxonomic lineage using the Ribosomal Database Project (RDP) classifier (v2.5) (46) with a Silva (release SSURef NR99 128 [47]) training set. Taxonomic summaries and taxonomic classifications were computed using QIIME v1.9.1 (48), and downstream analyses were done with in-house Perl and R scripts. Community richness and diversity were estimated based on 16S rRNA OTUs and data rarified to 14,900 sequences. Rarefaction analysis of each sample was performed 10 times to improve robustness. Community richness was estimated using the Chao1 species estimator (49). Community diversity was estimated using the Simpson index (50). For reasons of comparability and better interpretation, obtained diversity indices were converted to “effective number of species” (51). Community richness and diversity metrics were calculated using QIIME v1.9.1 (48). Statistical significance of diversity and richness differences was assessed using Student’s t test.
Quality filtering as recommended elsewhere (52) was performed prior to detailed analysis of community compositions: only OTUs with at least 0.05% relative abundance in at least 3 samples were retained. Microbial community compositions were compared based on principles of compositional data analysis (CoDA) (53). Dissimilarities between communities were expressed using the proportionality metric ϕ (54) as calculated using the R package “propr” (55). Obtained dissimilarities were used to cluster microbial communities by neighbor-joining tree estimation as implemented in the R package “phangorn” (56). Resulting tree topologies were visualized using the R packages “APE” (57) and “ggtree” (58). Differences between communities of setup/treatment subsets were explored using nonmetric multidimensional scaling (nMDS) as implemented in the R package “vegan” (59). Marginal effects of setup factors on community dissimilarities were tested using permutational multivariate analysis of variance (PERMANOVA) as implemented in the “adonis2” function of the R package “vegan” (59).
Differential abundance analyses of amplicon data.Differential abundance analysis of OTUs was performed using the “analysis of composition of microbiomes” (ANCOM) (60) framework. The ANCOM analysis was executed using code version 2.0 (https://sites.google.com/site/siddharthamandal1985/research) within the R environment. Where appropriate, differential abundances were adjusted (“adj.formula” option within ANCOM code) to account for differences between dilbit blends and setups with and without dispersant. ANCOM analyses were performed using a significance level of 0.05 and a 0.9 threshold for the W statistic. Genera that were differentially abundant in multiple-comparison analyses were identified and visualized using the R package “UpSet” (61).
Metagenomic and metatranscriptomic analyses of community composition.Metagenomic and metatranscriptomic sequencing was performed as described previously (29). Metagenome sequencing libraries were controlled for quality as described previously (29) and coassembled with a previous metagenomic data set of oil-degrading microcosms (NCBI Sequence Read Archive [SRA] no. SRP079000) (29) using Megahit v.1.1.2 (62) on a 3-Terabyte “Compute Canada” compute node. Gene coding sequences were predicted using MetaGeneMark v1.0 (63). Quality-controlled reads for both metagenomic and metatranscriptomic sequence libraries were mapped onto the assembly as described previously (29) to obtain an average read coverage value for each gene for each data type, i.e., metagenomic and metatranscriptomic. Community composition (metagenomes) and active fractions (metatranscriptomes) were determined using the single-copy rpoB gene universally conserved in prokaryotes (64). A reference database was created by retrieving all rpoB amino acid gene sequences contained in the Integrated Microbial Genomes & Microbiomes (IMG/M) (65) database (retrieved on 11 April 2017). Retrieved sequences were consolidated using cd-hit (66) and a 99% identity threshold. All predicted genes from the metagenomic coassembly were subjected to a BLAST search against this rpoB amino acid sequence reference database (available upon request), and all hits with an E value of <1e−10 were initially considered potential rpoB genes (in total, 7,236 genes). Of these potential rpoB genes, only genes featuring a BLAST score ratio (BSR) (67) of at least 0.3 to one of the rpoB reference sequences were retained during a second filtering step; this yielded a total of 1,088 sequences assumed to be true rpoB genes. The metagenomic and metatranscriptomic read coverage values of these genes were extracted from the computed abundance coverage values described above and are provided in Data Sets S15 and S16 in the supplemental material. Metagenomic and metatranscriptomic community dissimilarities were calculated and analyzed using the proportionality metric ϕ and nMDS as described above for 16S rRNA amplicon data.
Gene abundance analysis of metagenomic data.Gene abundance analysis was performed based on metagenomic data. The analysis was restricted to metagenomes with at least 3,000,000 reads; 95% of the metagenomic data sets fulfilled this criterion. Gene abundance analysis was focused on gene function. To this end, counts-per-million (cpm) values of genes were aggregated based on their associated Kyoto Encyclopedia of Genes and Genomes (KEGG) entry (68) and using the R package “dplyr” (69). Genes without associated KEGG entries were considered genes of unknown function. Aggregated gene functions with less than 1 cpm in fewer than 3 metagenomic data sets (each representing one microcosm sample) were removed from the analysis. Genes annotated with KEGG entries associated with hydrocarbon degradation pathways (K00496, K00529, K05297, K13953, K13954, K00121, K04072, K03738, K00114, K17760, K00128, K03380, K03379, K00462, K00446, K07104, K05549, K05550, K05784, K00154, and K05882) and at the same time being taxonomically classified as originating from typical hydrocarbon degrader genera (Polaribacter, Pseudomonas, Pseudoalteromonas, Marinobacter, Oleispira, Colwellia, Cycloclasticus, Thalassolituus, Alteromonas, Alcanivorax, Oceaniserpentilla, Oleibacter, and Oleiphilus) were assumed to be associated with hydrocarbon degradation. Abundances of hydrocarbon degradation genes were visualized using heat maps via the R package “ggplot2” (70). Differential abundance analysis of gene functions was performed using the R package “edgeR” (71). In detail, due to the large variation between our metagenomic data sets, cpm counts of gene function were used without further normalization. Differential abundance of gene function was determined using gene-wise negative binomial generalized linear models with quasilikelihood tests and a false-discovery rate (FDR) of 0.05. Specifically, we tested for differentially abundant gene functions between treatments with and without added Corexit 9500A while controlling for variation caused by the addition of different dilbit types by using a block design.
Data availability.All raw sequence reads generated for this study have been submitted to NCBI's SRA and are available under the accession number SRP152554 within the BioProject PRJNA450643. The metagenomic assembly is available under the BioSample accession number SAMN10986745 under the same BioProject. Metagenomic and metatranscriptomic read coverage tables of rpoB genes from microcosms are provided as Data Sets S15 and S16 in the supplemental material. The bioinformatics scripts used, as well as OTU count tables, are available upon request.
ACKNOWLEDGMENTS
We thank Cynthia Wright and Sophia Johannessen from the Institute of Ocean Sciences, Fisheries and Oceans Canada, in Sidney, BC, for their important role in consolidating the missions. We thank Peter Thamer, Brian Robinson, and Scott Ryan for their valuable contributions to this work in the field and in the laboratory. We also thank Gary Wohlgeschaffen, Claire McIntyre, and Graeme Soper for their participation in performing liquid-liquid extractions and Christine Maynard for her support with total nucleic acid extractions. We also thank the Canadian Coast Guard and specifically the crews of the CCGS John P. Tully and CCGS W.E. Ricker for their participation and assistance during field work. We acknowledge Compute Canada (www.computecanada.ca) for access to the University of Waterloo High Performance Computing (HPC) infrastructure (GP4/Graham system).
We declare no conflict of interest.
FOOTNOTES
- Received 11 January 2019.
- Accepted 1 March 2019.
- Accepted manuscript posted online 8 March 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.00086-19.
- Copyright © 2019 American Society for Microbiology.
