Native freshwater lake microbial community response to an in situ experimental dilbit spill

Abstract

With the increase in crude oil transport throughout Canada, the potential for spills into freshwater ecosystems has increased and additional research is needed in these sensitive environments. Large enclosures erected in a lake were used as mesocosms for this controlled experimental dilbit (diluted bitumen) spill under ambient environmental conditions. The microbial response to dilbit, the efficacy of standard remediation protocols on different shoreline types commonly found in Canadian freshwater lakes, including a testing of a shoreline washing agent were all evaluated. We found that the native microbial community did not undergo any significant shifts in composition after exposure to dilbit or the ensuing remediation treatments. Regardless of the treatment, sample type (soil, sediment, or water), or type of associated shoreline, the community remained relatively consistent over a 3-month monitoring period. Following this, metagenomic analysis of polycyclic aromatic and alkane hydrocarbon degradation mechanisms also showed that while many key genes identified in PAH and alkane biodegradation were present, their abundance did not change significantly over the course of the experiment. These results showed that the native microbial community present in a pristine freshwater lake has the prerequisite mechanisms for hydrocarbon degradation in place, and combined with standard remediation practices in use in Canada, has the genetic potential and resilience to potentially undertake bioremediation.

A large-scale controlled experimental dilbit spill into a pristine Canadian freshwater lake showed the resiliency and bioremediation potential of the native microbial community.

Introduction

Assessing the response of a pristine remote environment in the event of a potentially damaging contamination, like a dilbit spill, is integral to preparations for addressing such an event before it occurs. With the increase in demand for dilbit or crude oil and its various refined products, the potential risk of an accidental release into an otherwise pristine environment during transport likewise increases (Dew et al. 2015, Alsaadi et al. 2018). Canadian production and transport of crude oil is typically in the form of dilbit, or diluted bitumen. Crude oils are chemically comprised of compounds referred to as SARA: Saturates (alkanes, naptha), Aromatics (mono- and polycyclic), Resins (large heterocyclic compounds, soluble in n-pentane), and Asphaltenes [largest and most complex heterocyclic compounds with molecular mass up to 1000 g mol⁻¹ (Brown et al. 2017), insoluble in n-pentane] (Yoon et al. 2009). The composition of bitumen has a higher resin and asphaltene content than conventional crude (up to 15% w/w) and can be semisolid to solid and resistant to flow (Yoon et al. 2009, Stoyanovich et al. 2019). Thus, to aid in transport this solid bitumen is diluted with lighter fractions of oil (diluent, commonly naphtha) to form a less viscous flowing dilbit (Yoon et al. 2009, POLARIS Applied Sciences 2013).

Standard protocols for dilbit spill remediation vary situationally, but are based primarily on initial physical removal of the contaminants (POLARIS Applied Sciences 2013). Volumes of released dilbit can vary from several to hundreds of thousands of litres, and its fate in the environment is heavily influenced by routes of spread into the soils, groundwater, lakes, and rivers (Redondo and Platonov 2009, Dalton and Jin 2010). A myriad of remediation technologies are regularly employed with varying degrees of success (reviewed in O’Brien et al. 2017, Gitipour et al. 2018, Ossai et al. 2020). Removing contaminated matter through physical means and transporting to a facility for decontamination and/or destruction, ‘Dig and Dump’ is the most common and most effective protocol utilized (O’Brien et al. 2017, Gitipour et al. 2018). However, its effectiveness can be confounded by the presence of aboveground and belowground waterways, which can transport the contaminants over large areas, increasing the cost of removal and decontamination, and in some cases rendering the contaminated material inaccessible to physical remediation methods (Prendergast and Gschwend 2014). Bioremediation, a remediation method, which utilizes the metabolic capabilities of certain microorganisms extant in the environment to remediate hydrocarbon contaminants in situ in inaccessible areas has been successfully employed for decades in marine and Arctic environments. Deployment of chemical remediation strategies aimed at the mitigation of dilbit landing on shorelines such as dispersants (e.g. Corexit 9500), shore washing agents (e.g. Corexit 9580A, used to return crude oils to the water surface for increased physical recovery efficiency; Fiocco et al. 1991), or biostimulation are used to varying degrees of success in marine bodies of water (Ossai et al. 2020). However, their effectiveness in a freshwater system has not been established. Furthermore, our understanding of physical remediation methods, bioremediation, and the fates of dilbit in freshwater environments is currently incomplete.

A famous example is the offshore drilling rig Deepwater Horizon, where ~3.19 million gallons of crude oil (Wade et al. 2016) was released into the Gulf of Mexico, much of which remained inaccessible to physical removal methods. A commercial dispersant, Corexit 9500 was controversially added on both the surface and in the subsurface to address released oil inaccessible to physical methods. Corexit 9500 dispersed the released oil and increased it’s bioavailability to marine hydrocarbon-degrading microorganisms and minimized the amount of crude oil reaching the shore line (Mason et al. 2014, Kleindienst et al. 2015).

Numerous studies have delved into the fates of dilbit components; laboratory and field tests into the fate of spilled dilbit showed that only a small fraction of the released dilbit could be recovered from the water surface (12%–20%), with the remainder dispersed or dissolved into the water column (Dew et al. 2015, Stoyanovich et al. 2019). These dissolved or dispersed hydrocarbons would then be inaccessible for physical removal and could be addressed using microbial methods. The microbial degradation of dilbit has been demonstrated in freshwater and marine environments (Deshpande et al. 2018, Schreiber et al. 2019, 2021). However, in the event of an accidental dilbit release in a freshwater system, the complex geography and fluvial energy may act to disperse dilbit into the water column. While this dispersal/dissolution may be beneficial under certain marine spills (e.g. Deepwater Horizon; Mason et al. 2014, Kleindienst et al. 2015) increased concentrations of dispersed hydrocarbons in potentially precious freshwater systems can be devastating (Ariana 2016). Despite the amount of research into the effects and microbial degradation potential in marine environments, precious little has been performed for freshwater environments (Bhattacharyya et al. 2003, Nyman et al. 2007, Deshpande et al. 2018).

Here, we present the first of two studies as part of the Freshwater Oil Spill Remediation Study (FOReSt), with dilbit as the contaminant. FOReSt was established to understand the impacts of a dilbit or crude oil spill on the native microbial community present in a pristine freshwater lake and assess the effectiveness of minimally invasive remediation practices. A series of large enclosures were erected in situ in a freshwater lake located in the International Institute for Sustainable Development Experimental Lakes Area (IISD-ELA) in the summer of 2019. Controlled releases of dilbit was undertaken within these enclosures to mimic ‘real-life’ accidental dilbit spills, as well as simulating a typical remediation response timeline of 3 days from oil release to physical dilbit removal (conservatively taking into account the time needed to detect, assess, and transport personnel and infrastructure to a spill site). Additionally, the influence of biostimulation as well as the use of the SC Corexit 9580A shore washing agent as remediation and recovery strategies, respectively, were also tested. By conducting a large scale in situ experimental dilbit spills, we can acquire a greater understanding of the changes that the native microbial community might undergo and the ability of the native microbial population present in a freshwater lake to respond to or remediate dilbit in the event of a real spill can be examined. A second FOReSt experiment performed in 2021 using Canadian heavy crude is described in Kharey et al. (2024), which follows a similar experimental design.

Materials and methods

Study site

Lake 260 of the IISD-ELA in Ontario, Canada was used. Lake 260 (−93.76738, 49.69887) is an oligotrophic freshwater lake located on Precambrian bedrock with a surface area of 32.4 hectares and maximum depth of 14.6 m (Cleugh and Hauser 1971, McCullough and Campbell 1993). Shoreline enclosures were erected in the summer of 2019 perpendicular to the lake to contain the lake water and adjacent shoreline (Fig. S1, Supporting Information). The rectangular enclosures consist of a 20-cm square floatation collar from which a polypropylene curtain is suspended. The curtain was sealed to the lake bottom using a continuous double row of sandbags to contain the lake water and underlying sediments continuous with the shoreline. Enclosures straddled the shore and incorporated ~3 m of land and 10 m of the water surface, for a total length of 13 m and a width of 5 m. An estimated 30 000 l of lake water and 15 sq m of shoreline was enclosed in each shoreline enclosure.

Sets of enclosures were deployed on two different shore types, an organic rich wetland (W) and a less organic rocky cobble shore (RC). In total, eight enclosures were used for each shore type with the following treatments: enhanced monitored natural remediation (eMNR; n = 3), ShoreCleaner (SC; Corexit 9580A, n = 3), and Reference (no oil controls, n = 2). Oiling of the enclosed water surface was performed using 1.5 l of weathered Canadian Cold Lake Blend dilbit applied by an experienced oil spill technician. Prior to being applied weathering of the dilbit was done through volatilization by leaving dilbit exposed to the elements for 36 h on the surface of lake water from Lake 260 in a stainless-steel evaporation pan (1.1 m diameter). Weathered dilbit was collected into glass jars from the surface of the water using slotted stainless-steel spoons and then applied to the water surface of relevant enclosures within 50 cm of the shoreline. The dilbit was then allowed to interact with the shoreline for 72 h, after which dilbit was removed using low pressure freshwater flushing and capture of free-floating product using absorptive media. This weathering and 72-h gap between dilbit application and removal was included to mimic a typical real-life dilbit spill response where a few days delay between dilbit release, detection, assessment, and finally removal is often encountered. Experimental treatments (i.e secondary remediation) were then applied 1 day after the oil was removed. The experimental treatments were as follows: (1) eMNR: nutrients added to the eMNR enclosures targeting the Redfield C:N:P ratio using 50 g of Scott’s Osmocote slow release aquatic fertilizer (Redfield 1934) with 87% of the added dilbit assumed to be carbon. (2) SC: Corexit 9850A SC was applied to the oiled areas of shorelines in designated enclosures. After 5 min of contact time, the shorelines in the SC enclosures were again rinsed with freshwater and any additional recovered dilbit was removed with absorptive media. (3) Reference: no dilbit was applied in these treatment control enclosures. These enclosures were flushed with freshwater in an identical manner to the treatment enclosures during primary recovery to mimic release of terrestrial organic carbon into the enclosure aquatic environment. (4) Lake Reference: had no enclosures erected, received no dilbit, and no primary flushing as these control samples were taken from outside the enclosures.

Samples for microbial analysis were taken from the soil, sediment, and water from inside each enclosure in triplicates, including triplicate samples taken from the lake outside of the enclosures, these samples are referred to as ‘Lake Reference’ samples. The sediment (∼50 ml, within 20 cm into the waterline) and soil (∼50 ml, within 20 cm of the waterline) samples were taken from the shore within the top 5 cm of the surface, while all water samples (1 l grab samples from the surface) were taken and filtered on-site using 0.22 µm polyethersulfone membranes. All samples were immediately frozen on dry ice and shipped to NRC-EME (Montreal, Canada) on dry ice. All samples were kept at −80°C until gDNA extraction. Geochemical assays for the anion SO₄²⁻ analyzed on a Dionex DX-600 Ion Chromatograph (ThermoFisher, MA, USA), Fe was analyzed using an iCAP-6300 ICP-OES (ThermoFisher), DIC using a LI-850 LI-COR CO2/H20 Gas Analyzer (LI-COR Environmental, NE, USA), DOC using a Shimadzu TOC-VCPH/CPN (Shimadzu, Kyoto, Japan), and pH using a Fisher Scientific Accumet pH meter (Fisher Scientific, NH, USA) as described by Stainton et al. (1977). All geochemical and hydrocarbon analyses were performed by the IISD-ELA research team in Ontario, Canada. Concentrations of polycyclic aromatic hydrocarbons (PAHs) and their alkylated analogues were only extracted from water samples and analyzed using a validated gas chromatograph mass spectrometry (GC-MS) procedure following protocols as described in Idowu et al. (2017) using an Agilent 7890 coupled to a 7000C triple quadrupole mass spectrophotometer fitted with an electron ionization source. All PAH analyses on collected water samples were performed at the Center for Oil and Gas Research and Development (COGRAD) at the University of Manitoba.

DNA extraction

Total DNA was extracted from all samples using a Qiagen PowerSoil Kit (for sediment and soil samples, 1 g) or Qiagen PowerWater kit (whole filters), using the manufacturer suggested protocols (Qiagen, Hilden, Germany). Samples were eluted in 100 µl of PCR-grade water and a 20-µl aliquot of non-normalized DNA was taken for direct down-stream applications. All extracted DNA was stored at –20°C until further processing.

16S rRNA gene amplicon sequencing

The 16S rRNA gene was amplified using primers 515F-Y (5′-GTGYCAGCMGCCGCGGTAA) and 926R (5′-CCGYCAATTYMTTTRAGTTT), which target the V4–V5 region (Pyke et al. 2023). The PCR reaction protocol was as follows: 12.5 µl of KAPA Hi-Fi polymerase, 0.15 µl each of forward and reverse primers, 0.625 µl of bovine serum albumin (BSA), 9.5 µl PCR-grade H2O, and ~5 ng of DNA template. Thermocycling conditions used for 16S rRNA amplification were as follows: 95°C for 3 min, followed by 25 cycles of 95°C for 2 s, 58°C for 20 s, 72°C for 15 s, followed by a final extension at 72°C for 5 min. Indexing PCR reactions were identical to 16S rRNA amplification, reducing to eight cycles and removing the BSA additive. All samples (208) were purified using NuceloMag magnetic bead purification (Macherey-Nagel, Düren Germany) following the manufacturer’s protocol, and final purified amplicons were eluted in 50 µl of EB Buffer (Qiagen). Amplicons from indexing PCR reactions were purified using the same method as the initial amplification and eluted in 50 µl of EB Buffer (Qiagen). All samples were quantified using PicoGreen (ThermoFisher) using the manufacturer’s protocol. Samples were then normalized using a Hamilton Liquid Handling Robot (Hamilton Company, NV, USA) and pooled into amplicon libraries for loading onto an Illumina MiSeq Next-Generation Sequencer.

Metagenomic analysis

Select samples were further analyzed using metagenomics to assess the PAH degradation potential of the native microbial community. Samples were prepared for metagenomic analysis using Nextera XT Tagmentation, following the manufacturer’s protocol. Tagmented samples were then normalized, pooled, and sent to Genome Quebec (Quebec, Canada) for sequencing using an Illiumina HiSeq sequencer. Of all processed metagenomic samples, only the water SC samples taken on day 91 for both wetland and rocky cobble could not be processed due to issues in DNA amplification.

Bioinformatic analysis

Preparation of sequencing reads [including quality checking, paired-end assembly, and amplicon sequence variant (ASV) clustering] were done using the rRNA short amplicon pipeline (AmpliconTagger v.1.3.0) of the National Research Council Canada (NRC, QC, Canada) (Tremblay et al. 2015, Yergeau et al. 2015). Taxonomy was assigned using the Ribosomal Database Project (Cole et al. 2014) with SILVA 128 (Quast et al. 2013) using a 100% clustering threshold yielding ASVs. We acknowledge the age of this database may yield out-dated taxonomic assignment and have made efforts to use updated nomenclature as suggested by Oren and Garrity (2021) where necessary. Bioinformatics analysis was performed using R bioinformatics software and the Taxa Profiler web application (https://jtremblay2.shinyapps.io/taxonomy/) NRC, QC, Canada). ANOSIM (analysis of similarity) and PERMANOVA (permutational multivariate analysis of variance) were performed using R bioinformatics software packages ‘anosim’ and ‘adonis2’, respectively. PCA ordination for multivariate biplots was performed using ‘prcomp’ ordination method. NMDS plots were generated using Bray–Curtis dissimilarity metric. All mentioned packages were performed in R within the package ‘vegan’ (Oksanen 2013).

Metagenomic methods towards gene prediction (Prodigal v2.6.3; Hyatt et al. 2010), mapping (BBMap v38.11; Bushnell 2014), and genomic/genetic function and taxonomy assignment (RPSBLAST (v2.10.1+); Camacho et al. 2009) were carried out through an in-house pipeline ShotgunMG described in detail elsewhere (Tremblay et al. 2017, Tremblay and Greer 2019, Pyke et al. 2023) (https://bitbucket. org/jtremblay514/nrc_pipeline_public/src/master/). Identification of taxa with key genes located in assembled contigs/bins was done following a list of key aromatic and alkane degradation genes identified by our group (Tremblay and Greer 2019). Further analysis of select aromatic and alkane degradation genes by searching the KEGG ortholog (ko) annotation identities for ‘nahA’, ‘alkB’, ‘naphthalene dioxygenase’, and ‘alkane monooxygenase’. The resulting annotations were manually curated eliminating false positive identities.

Results

PAH and geo-chemistry

Across all dilbit applied enclosures, the concentration of total PAHs in water samples peaked on day 1 or day 20 postoil removal and returned to preoiling concentrations by day 66 (Fig. 1). The highest concentrations were measured in the eMNR and SC enclosures. eMNR in wetland peaked on day 20 postoil removal (2 ring = 2264.74 ng/l, 3 ring = 2121.75 ng/l, 4 ring = 3140.42 ng/l, 4+ ring = 21.04 ng/l, and total PAH = 7547.96 ng/l), and eMNR in rocky cobble PAHs peaked on day 1 postoil removal (2 ring = 1386.90 ng/l, 3 ring = 4976.29 ng/l, 4 ring = 2922.69 ng/l, 4+ ring = 32.28 ng/l, and total PAH = 9318.18 ng/l). Total PAH concentrations on day 66 in wetland and rocky cobble eMNR enclosures were <145 ng/l and <80 ng/l, respectively. In SC enclosures PAH concentrations peaked on day 20 in wetland shore enclosures (2 ring = 2756.44 ng/l, 3 ring = 2621.48 ng/l, 4 ring = 3958.94 ng/l, 4+ ring = 24.26 ng/l, and total PAH = 9361.12 ng/l) and day 6 in rocky cobble enclosures (2 ring = 5085.23 ng/l, 3 ring = 3128.89 ng/l, 4 ring = 2667.25 ng/l, 4+ ring = 35.15 ng/l, and total PAH = 10916.54 ng/l). By day 66 total PAH concentrations in wetland and rocky cobble SC enclosures were ∼145 ng/l and ∼79 ng/l, respectively. Slightly elevated concentrations of PAHs were also seen in the Reference and Lake samples during this time and followed the same downward trend in the dilbit-added enclosures. Forest fire activity in the area during that time may have been the source of this spike in PAHs (V. Palace, IISD-ELA, personal communication, 2024). Recovery of the released dilbit using standard physical remediation methods of freshwater flushing and recovery of free-floating products with oleophilic pads was low, ~10% by weight, which increased to 14% after repeated application of SC in the SC enclosures (V. Palace, IISD-ELA, personal communication, 2021). It should be noted that no attempt was made to scrub oil from vegetation or substrate as we were attempting to study minimally invasive secondary remediation. From the standpoint of oil recovery, the application of SC was not very effective on organic wetland substrates but was, not surprisingly, more effective on rocky shorelines (V. Palace, 2024, personal communication).

Figure 1.

Bubble plot of total PAH concentrations measured during the monitoring period grouped by number of aromatic rings (2–4+ rings and total PAHs) in wetland and rocky coble shorelines. Measurements began prior to dilbit addition (‘pre’), 1-day postdilbit removal, and continuing to the final timepoint 87-days postdilbit removal.

Geochemical analysis of collected wetland and rocky cobble water samples showed no differences between Reference (no dilbit control) enclosures and eMNR or SC treatment enclosures (Figs S2 and S3, Supporting Information). Standard deviations of replicate measurements (n = 3) encompass any trends across enclosures over the monitoring period.

Microbial community analysis

The microbial community composition of wetland soil samples over the ~90 days of monitoring remained relatively consistent, with strong clustering determined by sample source as soil, sediment, or water, rather than other experimental variables such as applied treatments, time, or the type of shoreline (Fig. 2). There is a proximity of clustering among soil and sediment samples, while water samples are distinct and cluster separately. ANOSIM analysis using Bray–Curtis dissimilarity metrics resulted in significant dissimilarity between sample types (R² = 0.9237, P = .001 at 999 permutations), and no significant dissimilarity when testing between treatments (R² = −0.02168, P = .974), and significant similarity when highlighting exposure day (R² = 0.02275, P = .03), or shoreline type (R² = 0.065, P = .001). PERMANOVA analysis with 999 permutations similarly resulted in no significant variability when analysing across treatment (R² = 0.0081 P = .871), but did reflect significant similarity with high explained variance among sample types (R² = 0.69701 P = .001), with low explained variance among exposure day (R² = 0.06349 P = .001), and shoreline type (R² = 0.03546 P = .002) groupings.

Figure 2.

Multivariate biplot of all 2019 FOReSt samples collected across all enclosures, shore types, and sample types over the monitoring period. Ellipses denoting 99% confidence were determined through the sample type variable. The 20 most relatively abundant taxa (g = Genus; OR = Order; and CL = Class) are shown as vectors denoting association between taxa and sample points. Vectors use top and right axes.

The microbial community composition of soil samples in both wetland and rocky cobble shoreline types over the ~90 days of monitoring remained relatively consistent (Fig. S4, Supporting Information). In the organics rich wetland shoreline type, a marked but nonsignificant decrease in the abundance of unidentifiable members of the order Acidobacteriales, and in particular Acidobacteriales family Subgroup 2 was seen under SC treatment and Reference controls, but less so in eMNR and Lake Reference enclosures. This trend was not seen in organics poor rocky cobble shoreline type samples. The dominant members of both the rocky cobble and wetland shoreline type enclosures are nearly identical as seen in Fig. S5 (Supporting Information), where the clustering of samples using Bray–Curtis dissimilarity shows they are closely related over the monitoring period. This trend was not seen in any other sample source nor experimental condition. Multivariate analysis shows a number of the most dominant taxa strongly associated with soil samples: Bradyrhizobium, Bryobacter, Candidatus Solibacter, unassigned members of the orders Burkholdariales and Acidobacteriales, members of the hgcI clade, and members of the Pir4 lineage are strongly associated with soil samples (Fig. 2).

Similar to trends seen in soil samples, no prevalent change in community composition was seen in sediment samples over the ∼90 days of monitoring, regardless of treatment or shoreline type (Fig. S6A, Supporting Information). This consistency of the 20 dominant taxa in the wetland and rocky cobble shoreline bacterial communities over time was reflected in the NMDS plot of the whole microbial community using Bray–Curtis dissimilarity (Fig. S6B and D, Supporting Information). The most abundant taxa present in the wetland sediment, were Kryptoniales, Bathyarchaeia, Sphingobacteriales, Spirochaeta, Bacteriodales, RBG-13–54–9, and SBR1031 (both members of Anaerolineae). Kryptoniales, Rickettsiales, Sphingobacteriales, and others were most dominant in rocky cobble samples albeit slightly differing in abundance compared to wetland. Some slight changes in the abundance of Bathyarchaeia and Acidobacteriales were seen in the preoiling time point. However, after this initial shift, the communities did not seem to change much over the ∼90-day monitoring period and are relatively similar in composition to the Reference and Lake Reference samples. The preoiling timepoints seemed to be more dissimilar, with the community becoming less dissimilar over time with no influence of treatments (Fig. S6D, Supporting Information). Members of the orders Kryptoniales and Sphingobacteriales, and members of the class Bathyarchaeia were associated with sediment communities using multivariate analysis (Fig. 2).

The most dominant taxa present in enclosed wetland waters changed in abundance over the monitoring period with a gradual decrease in Polynucleobacter and Sediminibacterium across all treatments (Fig. S7A, Supporting Information). Relative to the Lake Reference samples, the microbial community present within the wetland enclosures initially increased in abundance of Polynucleobacter, and members of the order Rickettsiales, and there was a drastic reduction in SAR11 clade and Cyanobium PCC-6307, prominent members in the Lake Reference samples and the rocky cobble water samples (Fig. S7C, Supporting Information). The most abundant taxon in water samples, ‘hgcI_clade’, was also detected in all sample types. Very little change occurred in the rocky cobble among the 20 most abundant taxa over time and relative to the Lake Reference samples. However, strong clustering of samples was observed over time when analyzing the entire microbial community (Fig. S6D, Supporting Information). Candidatus Planktophilia, Cyanbobium PCC-6307, Polynucleobacter, Sediminibacterium, and members of the orders Chloroplast, Rickettsiales, and SAR11 clade were associated with water samples using multivariate analysis (Fig. 2).

Ferruginibacter and unassigned members of SBR1031 were found to be negatively associated with sediment and water samples, respectively, and were not strongly associated with any sample type (Fig. 2).

Metagenomic analysis of freshwater exposed to dilbit

The metagenomic analyses presented are focused on water samples, where oiling predominantly occurred and where changes in the microbial community were observed.

Genes associated with aromatics degradation

Aromatic degradation gene abundances in the enclosures erected on wetland and rocky cobble shoreline types remained relatively stable over the monitoring period (Fig. 3). These genes were detected across all sample types and in samples taken before oil-addition. Key genes involved in aromatic ring attack like catechol 1,2 dioxygenase (catA), catechol 2,3 dioxygenase (catE/dmpB/xylE), toluate 1,2 dioxygenase (benA/xylX), and other key genes were detected across all collected samples. Genes associated with catechol attack (catechol 1,2 and 2,3 dioxygenases) and benzoate/toluate dioxygenases were the most abundant aromatic degradation-associated genes. In the water samples, abundances increased in some genes associated with the metabolism of potential PAH degradation metabolites (catechol 2,3 dioxygenase, and putA coniferyl aldehyde dehydrogenase). Relative abundances of all measured genes were higher in the soil and sediment than they were in the water matrix.

Figure 3.

Bubble plots showing relative abundance of genes involved in the degradation of aromatics found in wetland and rocky cobble shore type samples in each treatment. Treatments, eMNR (enhanced Monitored Recovery), SC (ShoreCleaner—Corexit 9580A), and no-oil controls [Ref (Reference) or LakeRef (Lake Reference)] over the 91-day monitoring period. Genes involved in aromatic degradation were seen throughout the monitoring period.

Assignments of nahA sequences extracted from our wetland (W) and rocky cobble (RC) dataset show that phylum level assignments are nearly equally distributed across Alphaproteobacteria (W = 13.8%–41%, RC = 15%–32%), Thermoplasmata–Methanomassiliicoccales (W = 8.6%–35%, RC = 19%–36%), Actinomycetota (formerly Actinobacteria) (W = 4.2%–27%), Betaproteobacteria (W = 3.1%–22%, RC = 9.3%–18%), Bacteroidia (W = 0.27%–20%), and Acidobacteriota (formerly Acidobacteria) (RC = 4.7%–31%) (Fig. S8A and B, Supporting Information). Abundances of nahA sequences assigned to the genus level were dominated by Limnohabitans (W = 4.8%–28%, RC = 3.2%–11%), Polynucleobacter (W = 0.7%–41%, RC = 0.3%–4.4%), Rhodoferax (W = 0.27%–16%, RC = 0.3%–3.1%), and Aquabacterium (W = 0.29%–9.6%, RC = 0.6%–4.8%) with no trends specific to any treatment (Fig. S8C and D, Supporting Information). The remaining 29%–82% of assigned nahA sequences were present on contigs unable to be assigned at the genus level. Sequence identification of ‘NULL’ refers to sequences, which were unable to be assigned taxonomy to the specified taxonomic rank of either phylum or genus.

Genes associated with alkane degradation

Medium to long chain alkanes, with carbon chain lengths from 14 to greater than 25, are present in dilbit and comprise around 30% of the dilbit content (POLARIS Applied Sciences 2013). Their presence in dilbit indicates that their degradation potential should also be considered alongside aromatic components.

Alkane degradation genes followed similar trends to the aromatic degradation genes, where the soil and sediment abundances remained static over the monitoring period, (Fig. 4). Key alkane degradation genes such as those involved in the initial activation of alkanes (alkB1 and alkB2, alkane 1 monooxygenase, and ladA, a long chain alkane monooxygenase) were found in appreciable abundance in predilbit timepoints. Additionally, there was a marked increase in the abundances of some late-stage alkane degradation associated genes, such as aor aldehyde ferredoxin, alcohol dehydrogenase, and adhP propane preferring alcohol dehydrogenase in the water samples with eMNR treatment. This trend was more pronounced in the rocky cobble enclosures with increases in those gene abundances in the SC treatment as well.

Figure 4.

Bubble plots showing relative abundance of genes involved in the degradation of alkanes found in wetland and rocky cobble shore type samples in each treatment. Treatments, eMNR (enhanced Monitored Recovery), SC (ShoreCleaner—Corexit 9580A), and no-oil controls [Ref (Reference) or LakeRef (Lake Reference)] over the 91-day monitoring period. Genes involved in alkane degradation were seen throughout the monitoring period.

Taxonomic assignments at the phylum level of the alkB sequences from the metagenomic dataset showed that Betaproteobacteria dominate (W = 13%–48%, RC = 40%–56%), Desulfobacterota (formerly Deltaproteobacteria) (W = 0.7%–32%), Gammaproteobacteria (W = 2%–43%), Chitinophagia (RC = 5%–30%), and Actinomycetota (Actinobacteria) (RC = 4%–20%) (Fig. S9A and B, Supporting Information). Members of Flavobacteriia were found to be most abundant in the eMNR (day 34, 19%) and found in Reference samples between 0% and 9%. No trends unique to treatments were seen in the alkB abundances, nor were any particular group unique within a particular treatment. Genus level assignments showed similar results in that there is no predominance of a certain group of genera to any specific experimental condition (Fig. S3, Supporting Information). Ideonella (W = 0%–35%), Pseudomonas (W = 0%–39%), Curvibiacter (W = 0.2%–31%), Rhodoferax (W = 0.4%–10%, RC = 0%–7.5%), Stylonychia (RC = 0.1%–9.9%), Emticicia (RC = 0.7%–6.4%), and Sediminibacterium (RC = 0%–8.1%) were the most abundant genera assigned to the genus level, with the remaining 42%–83% of sequences unassigned to the genus level (Fig. S9C and D, Supporting Information). Ideonella was enriched in wetland eMNR enclosures on day 34, and otherwise found in the reference enclosure on the same day.

Discussion

The FOReSt study was designed to increase understanding of the effects a dilbit spill on the microbial communities present in the soil, sediment, and water of a pristine freshwater lake and to examine the effects of minimally invasive dilbit spill remediation. Furthering our understanding of the fates of dilbit PAH components and the efficacy of a typical oil removal response including testing biostimulation and oil recovery type remediation methods.

In this study, the concentrations of PAH compounds ranging from 2 to 4+ rings decreased to background concentrations within 66-days postdilbit removal. This result is in line with a second FOReSt study described in Kharey et al. (2024), where PAH components of a Canadian heavy crude oil returned to background concentrations within 14–73 days. Trends in the decrease of PAHs over time are also similar, where the wetland concentrations decreased over a longer period of time in this 2019 FOReSt and in the 2021 heavy crude oil FOReSt (Kharey et al. 2024). The rocky cobble concentrations peaked and decreased at earlier timepoints, compared to wetland. Additionally, the concentrations of PAH compounds measured here were nearly 50-fold higher than measured in the heavy crude oil contaminated FOReSt study (Kharey et al. 2024). To our knowledge, differences in residual PAHs between dilbit and heavy crude oil have not been reported previously. It must be made clear that the 2021 FOReSt experiment presented in Kharey et al. (2024) was performed in shoreline enclosures identical to those used in this study but located in a different part of Lake 260. As such the possibility of PAH carry over, or cross enclosure contamination is negligible.

Studies on the fate of dilbit and other crude oils in water under laboratory or field conditions showed that more crude oil components were dissolved or dispersed into the water column through wave action, UV degradation, or other environmental factors (Lee et al. 2013, Dew et al. 2015, Alsaadi et al. 2018, Stoyanovich et al. 2019). However, wave action was negligible within the enclosures due to the low energy environment of the study lake and the additional damping action of the surface enclosure structures that extended 20 cm above the water surface (V. Palace, IISD-ELA, 2021, personal communication). Dispersion, dissolution, photooxidation, and microbial metabolism may all be in-part responsible for the decrease in PAH concentrations. Identifying the catalyst for PAH loss over time is confounded by the environmental conditions of this experiment, where environmental variables (volume of enclosure waters, sunlight levels, wind speeds, diurnal temperature fluctuations, heterogeneous shore structures, vegetation, and so on) can contribute to perceived and/or actual hydrocarbon loss. This closely reflects the conditions expected in a real-world freshwater dilbit release, suggesting that spills of small quantities of dilbit over relatively small surfaces (i.e. <2 l/20 m², this study) and physically recovered may be effectively remediated within a few months.

Dilbit–induced changes in the microbial community in freshwater soils, sediments, and water

The microbial community was strongly influenced by the phase of the freshwater lake, with little to no influence from the dilbit, or the applied remediation treatments (Fig. 2). Over the 90-day monitoring period no indicators of microbial stress were seen to alter the microbial community in the soil, sediment, or water, with minor exceptions.

Within the soil samples, the presence of putative hydrocarbon degraders like Burkholdariales (Jeon et al. 2004, Lünsmann et al. 2016), Sphingobacteriales (Pagé et al. 2015, Zhong et al. 2017), and unidentifiable members of the order Acidobacteriales (Kadnikov et al. 2013, Bourceret et al. 2016, Cecotti et al. 2018) suggest potential hydrocarbon biodegradation occurring in soil samples. Other taxa seen in these samples, such as Candidatus Solibacter (Puranik et al. 2016) and Bradyrhizobium (Puranik et al. 2016, Jin et al. 2022), are common soil inhabiting bacteria involved in nutrient cycling in the rhizosphere and not directly implicated in hydrocarbon biodegradation.

The presence of some putative hydrocarbon degrading taxa in the sediments such as Sphingobacteriales, Bacteriodales, and Anaerolineae, which have been described in some PAH-contaminated soils and sediments, may indicate the potential for natural attenuation of residual PAH and aliphatic compounds (Kamarisima et al. 2019, Zhong et al. 2017, Gieg et al. 2014). These taxa were not abundant in the 2021 FOReSt soil or sediment samples, which were taken during similar times in their respective sampling year, suggesting that the freshwater microbial community has a degree of heterogeneity over time and/or space (Kharey et al. 2024). The sediment community was the most resistant to change among the freshwater lake communities. Study of weathered dilbit penetration in different shore types indicated <5 cm of penetration and a low retention volume (<100 l/m³) in rocky sediment types (POLARIS Applied Sciences 2013). Sampling of sediments was done within the upper 5 cm of the sediment, and it was noted that dilbit did not accumulate on the shore during the experiment (V. Palace. 2021, personal communication). The nature of the sediment matrix may have contributed to the lack of changes in the microbial community, where an influx of nutrients and/or potential hydrocarbon substrates may have been limited (Verrhiest et al. 2002, Brion and Pelletier 2005). The microbial community was the most stable across all sampled areas and the only areas where anaerobic methanogen-associated taxa (Anaerolineae) were detected, was where the lack of oxygen diffusion and low redox potential are generally found. Due to the physical barriers required to eliminate mass transfer from the enclosures into/out of the lake, it was noted that the wave action inside the enclosures was nearly entirely eliminated (V. Palace, IISD-ELA, personal communication). The lack of this water activity may have influenced the interplay between the different matrices of the lake inside the enclosures.

Similar to the soil community, the microbial community present in the water samples changed slightly over the monitoring period. The most prominent change was seen in the wetland enclosure group where the microbial communities within the enclosure deviated from the community in the lake proper. This shows that the enclosures themselves impacted the water microbial community. From these enclosure-wide changes the water community as a whole was more susceptible to change over the monitoring period compared to the soil and sediment communities. The presence of the same putative hydrocarbon taxa in the water, soil, and sediment suggests that hydrocarbon degradation is a metabolic function, i.e. shared across the different phase of the freshwater lake (Deshpande et al. 2018). This enclosure mediated shift in microbial community may stem from the removal of limnal flow, whereby stagnating organic compounds such as humic acids, lignins, cellulose, and so on, which can share degradation mechanisms, may confound possible dilbit hydrocarbon stress. This may explain the pronounced changes in the organics-rich wetland shoreline relative to the organics-poor rocky cobble shoreline. The most abundant taxon in water, the ‘hgcI_clade’ was detected in all other sample types and has been detected in methane seeps (Zemskaya et al. 2015). However, despite the introduction of dilbit and the applied treatments to the water surface, the microbial community did not reflect stress conditions. The hypothesis that the residual dilbit and applied treatments in the experimental enclosures would reflect a metabolic stress as an increased abundance of putative hydrocarbon degrading taxa was not seen in the microbial community.

Assessing that the shoreline type makes a difference on the ability of the lake communities’ potential response to a dilbit spill can be summarized by Fig. 2 where the all the samples are clustering together, exclusively by the source of the samples, rather than by the influence of any experimental variable (time or treatment). The compositions of the microbial communities in the soil, sediment, and water between these two separated, yet connected, areas by way of limnal flux throughout the lake/river system is related. The most dominant taxa are consistent between the two shoreline types while slight variation in dominance is prevalent. This lack of trend across the different microbial diversities of the lake puts the previously discussed trends within each lake shoreline type into perspective: that the microbial community was most likely influenced by the static matrices of the lake (shoreline sediments and soil) and that the residual dilbit, which remained after physical remediation of a small volume of supplied dilbit was well-tolerated by the autochthonous freshwater microbial community.

These trends of little to no change in the microbial community in the soils, sediments, and water were also seen in the 2021 heavy crude oil FOReSt study (Kharey et al. 2024). The applied biostimulation treatment, like that described here, had no discernable effect on the microbial community, and putative hydrocarbon degrading taxa were detected.

From the community data it is not apparent that the eMNR or SC treatments had an effect on increasing the abundance of relevant oil degrading taxa, however, community data cannot be directly correlated to metabolic activity. As discussed earlier, some shifts in the microbial community present in the water samples were seen in the eMNR and SC enclosures, that were not observed in the soil and sediment, such as Polynucleobacter and SAR11 clade. These changes in abundance may suggest influences of seasonality or other environmental factors. Many relevant oil degrading taxa seen in the soil and sediment samples are also present in these water samples, albeit at lower abundances, such as Acidobacteriales, Burkholderiales, and so on. The presence of these taxa in the water samples, the primary location of oiling suggests that residual PAHs have the potential to be biodegraded.

Presence of key aromatic and alkane degradation genes in dilbit-contaminated wetland or rocky cobble shoreline types

Detection of key genes involved in aromatic ring attack like catechol 1,2 dioxygenase (catA), catechol 2,3 dioxygenase (catE/dmpB/xylE), toluate 1,2 dioxygenase (benA/xylX), and key genes used as diagnostic markers for PAH degradation such as naphthalene dioxygenase (nahA/nagA; Bosch et al. 1999, Dionisi et al. 2004), and others, showed that the microbial community possesses the necessary mechanisms for aromatics biodegradation. Genes associated with catechol attack (catechol 1,2 and 2,3 dioxygenases) and benzoate/toluate dioxygenases are the most abundant aromatic degradation associated genes, likely due to the abundance of aromatics in complex natural structures (lignin, humic acids) and as a downstream product of PAH degradation (De Haan 1976, Szabó et al. 2007). Pyrogenic PAHs, those originating from combustion events such as forest fires contribute as natural sources of aromatic hydrocarbon degradation activity (Campos and Abrantes 2021, Yang et al. 2022). The ubiquity of aromatics in the lake environment likely selects for this activity in otherwise pristine environments. These key hydrocarbon degradation genes were seen across all sample types and in samples taken before oil-addition, establishing that this pristine environment has the necessary genetic potential to confront a hydrocarbon release event.

The applied treatments, as was seen in the microbial community composition analysis, had no discernable affect on the relative abundance of any aromatic degradation gene in the soil and sediment samples; any changes in abundance in the treatment enclosures were seen, or were closely mirrored by the reference no-oil enclosure and the lake reference samples. However, in the water samples there was an approximate log-fold increase in the abundances of nearly all assayed genes in the organics-poor rocky cobble dilbit and eMNR amended samples. Interestingly, the use of SC 9580A had a similar effect on gene abundance to the biostimulation treatment, which suggests that the presence of residual dilbit rather than this washing treatment stimulated the increase in rocky cobble samples, while no increase was seen in the wetland samples. This is an unexpected result as the addition of fertilizer to oil contaminated samples has been seen to stimulate degradation activity in wetland environments (Greer et al. 2003), while the lack of influence using chemical additives has also been observed (Nyman et al. 2007). It should also be noted that the abundances reached on the dilbit stimulated rocky cobble are similar to the apparently static abundance seen in the wetland samples. As was discussed above, neither biostimulation nor SC were seen to have influenced the microbial community composition. It seems that the influx of aromatic substrates from dilbit, where substrate analogues may be more readily available from the wetland shore type in the form of potential plant and ligneous substrate sources, led to an increased abundance in aromatic degradation mechanisms, while having no discernible affect on the taxonomic makeup.

Alkane degradation genes such as alkB1 and alkB2, alkane 1 monooxygenase, and ladA are key genes in the activation of alkane compounds for aerobic degradation (Nie et al. 2014). While hydrocarbon degradation has been predominantly studied in marine systems, alkane degradation and the detection of alkB genes have been shown in freshwater systems (Masuda et al. 2014, Reid et al. 2018). These genes were found to be less abundant than downstream fatty acid degradation genes across all enclosures and in predilbit timepoints. The ubiquity of downstream alkane degradation genes, such as fatty acid degradation, reflects a more prevalent metabolism than initial alkane chain metabolism (Fujita et al. 2007, Parsons et al. 2014). Similar to the trends seen with the aromatic genes, there was an increase in the abundance of nearly all assayed alkane degradation genes in the treatment enclosures, but not in the controls. Again, this suggests that the addition of dilbit, rather the residual dilbit remaining in the enclosure after removal, likely had a stimulatory affect on alkane degradation gene harbouring microbes. Further quantification of these alkane and afore-discussed aromatic degradation genes was confounded by the vast diversity of these sequences found in our study site (data not shown). The presence of these genes establishes the potential for alkane degradation to occur in this pristine freshwater lake. The presence of these genes in preoiling timepoints, shows that the native microbial community present has members with that metabolic pathway and can be utilized in the event of a dilbit spill.

Tracing the identity of key PAH and alkane degradation gene harbouring microbes

Metagenomic assignments of the nahA gene on assembled contigs highlights the degradation potential for specific taxa. Aromatic hydrocarbon degradation prominent genera included Limnohabitans, Rhodoferax, Polynucleobacter, Aquabacterium, and members of Alphaproteobacteria. Rhodoferax (Aburto and Peimbert 2011) and unassigned members of Alphaproteobacteria (Cederwall et al. 2020) are putative PAH degrading taxa with the exception of Polynucleobacter, which is known for humic acid degradation (Hahn et al. 2012), Aquabacterium which has been identified with alkane degradation mechanisms (Masuda et al. 2014) and Limnohabitans a nondescript freshwater microbe (Jezberová et al. 2017). The ubiquity of transferable genetic elements (i.e. plasmids) with hydrocarbon degradation genes is very common and may explain why taxa such as Limnohabitans and Aquabacterium may harbour these or multiple degradation mechanisms (Whyte et al. 1997, Zhou et al. 2001). The most abundant nahA sequence was annotated to the Archaeal order Methanomassiliicoccales (8.6%–35% relative abundance), the sole assigned member of Thermoplasmata. The presence of this strict anaerobic hydrogenotrophic methanogen (Andrea et al. 2016) had not been detected during genomic assays discussed above, and its annotation with a nah mechanism, a distinctly aerobic mechanism, is paradoxical. Searching the NCBI gene directory for Methanomassiliicoccales and ‘nahA’, ‘naphthalene’, or ‘dioxygenase’ yielded no results, suggesting that the annotation of nahA may either be novel in this study or a misannotation/assignment. While the taxonomic abundance of these genes was seen to be incredibly diverse, neither the treatments, nor the influence of the enclosures was apparent at this genetic level, as was seen in the 16S rRNA gene community data discussed earlier.

Prominent alkB harbouring genera were identified as Ideonella, Rhodoferax, Stylonychia, Sediminibacterium, as well as members of Beta- and Gammaproteobacteria, and Desulfobacterota (formerly Deltaproteobacteria). Of these identified genera, only Rhodoferax has been described as associated with hydrocarbon degradation communities (Aburto and Peimbert 2011, Aburto-Medina et al. 2012, Martin et al. 2012). Rhodoferax was seen to have both the nahA and alkB genes; while it is not known if these genes are located on a plasmid or the chromosome, multiple degradation mechanisms have been characterized on a single plasmid in other bacterial species (Whyte et al. 1997). The presence of Ideonella, a plastic degradation-associated microbe is interesting as it was isolated on polyethylene terephthalate plastic, which does not have alkyl side chains. Ideonella is associated with, but not characterized as, an alkane degrader (Giebler et al. 2013), but species have been described in monoaromatic degradation (Bedics et al. 2022). Interestingly, Stylonychia a eukaryotic freshwater ciliate, was found in greater than 10% total alkB abundance across all samples. This is not a novel instance as Stylonychia has been shown to have an alkane degradation mechanism and a PAH-specific CYP450 (Cid et al. 2017, Yim et al. 2017). It is crucial to note the presence of misannotated alkB sequences due to the unfortunate similarity in gene symbol to a eukaryotic dioxygenase alkB involved in DNA/RNA demethylation (Westbye et al. 2008, Cid et al. 2017). Misannotations of alkB in sequences from databases to which the sequence identities were made were rife in this dataset and were manually removed from these analyses. Extracting sequences from the dataset was done by searching for ‘alkB’ and filtered for those with the function annotation of ‘monooxygenase’.

Many of these genera are not typically regarded as PAH or alkane degraders, but the analysis presented here suggests that in the freshwater lake assayed they may fulfill that role. The taxonomic makeup of this key aromatic and aliphatic degradation mechanism was seen to be very diverse, with no singular taxonomic group occupying that metabolic niche. Further pangenomic analysis into the degradation abilities of taxa, which were consistently identified with nahA and alkB mechanisms such as those of Polynucleobacer, Limnohabitans, and Aquabacterium is currently underway.

Conclusions

This project was the first to undertake a large-scale experiment to monitor the response of the indigenous microbial community to spilled diluted bitumen and minimally invasive remediation methods in situ in a pristine freshwater lake. While major changes in the microbial community were not seen over the sampling time, the presence of putative hydrocarbon degrading taxa, along with the ubiquitous presence of key alkane and PAH degradation mechanisms and pathways across different lake matrices (soil, sediment, and water) established the potential of the native microbial community in a pristine freshwater lake to biodegrade dilbit-derived contaminants.

Containment of spilled dilbit, followed by physical removal is an effective strategy in the remediation of contaminated freshwater limnal systems. The reduction of PAHs to precontamination levels after 66 days reflects the efficacy of the standard oil-removal practices and the ability of the PAH and alkane degradation mechanisms present in the native microbial communities in the event of a dilbit release. The slight increase in the abundance of genes associated with hydrocarbon biodegradation in the dilbit-added enclosures suggests that these mechanisms may be stimulated. The use of eMNR or SC as treatments did not have an obvious impact on the microbial community nor the genetic assemblage of PAH and alkane degradation mechanisms as applied here. However, their efficacy as remediation strategies or supplements should not be disregarded, as biostimulation and the use of chemical additives are effective and widely used bioremediation strategies. The generally static trends seen throughout this experiment are most likely attributed to the lower concentrations of residual PAHs persisting in the enclosures, as well as the scale of the enclosures and inherent heterogeneity of the enclosure environment, made more complex with environmental factors and variables, including diurnal changes in temperature, flux of the water column, predation, presence of other complex organic compounds such as humic acids, lignins, and so on. The deployment of the enclosures themselves was a factor in microbial community shifts. The natural presence of compounds that are similar in structure to alkanes and aromatics (humic acids, lignin, and so on) may have contributed to the observed microbial community trends. As such, the influence of residual hydrocarbons was minimal to nonexistent in the soil and sediment matrices, while community shifts were seen in the water samples, where bioavailable PAHs and alkanes would be expected following an over-water oil release.

The use of the IISD-ELA provided a unique opportunity to explore larger scale, in situ experiments where phenomena like oil spills can be studied under controlled conditions outside of a real-world ecological disaster. These data show that further research into understanding the effect of scale and environmental conditions on hydrocarbon degradation in freshwater systems is needed. Further work characterizing the effects of the introduced contaminants and applied treatments to the indigenous microbial community dynamics and resiliency are currently being undertaken.

Author contributions

Gurpreet S. Kharey (Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing), Vince Palace (Conceptualization, Funding acquisition, Methodology, Resources), Lyle Whyte (Funding acquisition, Project administration, Supervision), and Charles W. Greer (Project administration, Resources, Supervision, Validation, Writing – review & editing)

Conflict of interest

The authors report no conflict of interest.

Funding

This work was supported by a Collaborative Research and Development (CRD) grant from the Natural Science and Engineering Research Council of Canada (NSERC) (#CRDPJ-532225) awarded to G. Tomy (University of Manitoba) and by a Genome Canada GAPP grant awarded to V Palace.

Data availability

Sequencing data generated for this project were deposited in the NCBI SRA portal under the BioProject PRJNA1079791. Metagenomic shotgun data are available under the SRA accession numbers SRR28094181–SRR28094201 (water), SRR28094950–SRR28094970 (lake sediment), and SRR28073467–SRR28073488 (lakeshore soil). 16S rRNA gene amplicon data are available under the same BioProject.