Introduction

Oil is frequently produced by water injection. Wells in many mature on-land reservoirs typically produce 5–10% oil and 90–95% water, with the water cut increasing over time (Planckaert 2005; Youssef et al. 2007). Microbially enhanced oil recovery (MEOR) is one of the technologies that may help increase the fraction of the recovered oil (Bryant and Burchfield 1989; Banat 1995; Li et al. 2002; McInerney et al. 2005). Microorganisms attached to oil are particularly interesting for MEOR applications. They may be present in core matrix pores, where oil is trapped in micro-droplets and where microbial activity may lead to oil degradation and methane production (Zengler et al. 1999; Head et al. 2003).

It has been shown that microorganisms may need to produce alcohols and/or biosurfactants to get attached to an oil–water interface (Abbasnezhad et al. 2008). On the other hand, surfactants, especially when combined with alcohols, are known to reduce interfacial tension between the hydrocarbon and aqueous phases and may in this way enhance oil recovery (Bryant and Burchfield 1989; Banat 1995; Li et al. 2002; McInerney et al. 2005). Cells attached to the oil–water interface are also known to behave like adsorbed colloidal particles. Thus, they may allow the application of an enhanced oil recovery mechanism, which does not depend on interfacial tension reduction, but instead depends on plugging of high permeability zones (Kang et al. 2008). It is also important to note that in some MEOR applications transportability of the microbes throughout the oil reservoir may be an important parameter, and in such cases reliance on the microbes associated with the aqueous phase may be required.

A number of methods have been used to study hydrophobic interactions of cells, for example those based on defining the degree of adhesion of cells to liquid hydrocarbons following a brief period of mixing; contact angle measurements of dried cell layers; and partitioning of bacteria in aqueous polymer two-phase systems (Rosenberg et al. 1980). A method known as MATH (microbial adhesion to hydrocarbons) has gained popularity due to its simplicity and acceptable level of accuracy, despite some disadvantages (Rosenberg 2006; van der Mei et al. 1995; Zoueki et al. 2010). It is based on defining the degree of adhesion of cells to liquid hydrocarbons following mixing (Rosenberg et al. 1980; Rosenberg 2006). In this study, we employed an idea similar to that of the MATH method in order to determine the compositions of the OC (microbial community associated with an oil phase) and the AC (microbial community associated with an aqueous phase). In our method, oil-attached cells firstly rise to the surface with the oil during gravitational separation (1×g), and secondly they are dislodged from the oil and pelleted by centrifugation at 104×g. To our knowledge, the use of this method has not been reported previously.

Materials and methods

Isolation of biomass

Samples of oil-containing produced water were collected in 1l Nalgene bottles from production wells in the Enermark field (near Medicine Hat, Alberta, Canada) on 12/01/2010 and 02/02/2010. The bottles were filled to the top to prevent oxygenation (Voordouw et al. 2009), delivered to the laboratory within five hours after filling and stored in an anaerobic hood (90% N2 and 10% CO2) at 20°C. The samples were separated into oil and aqueous phases in the anaerobic hood in a separatory funnel (gravitational separation, 1×g). Following standing overnight, the aqueous phase was drained off and biomass associated with it was collected through centrifugation at 25,000×g for 20 min in 250 ml centrifuge bottles. The pellets were re-suspended in 1 ml of sterile synthetic brine (90 mM NaCl; 1 mM KCl; 1 mM MgCl2; 2 mM CaCl2; 10 mM NaHCO3; pH 7.5) by agitating for 50–60 s with a Vortex mixer, transferred to 1.5 ml microcentrifuge tubes and centrifuged at 17,000×g for 10 min. The supernatant was removed and collected biomass was subjected to DNA extraction. The separatory funnel containing the remaining oil phase was replenished with 1–1.5 l of sterile synthetic brine, manually shaken for 2 min and allowed to stand for at least 24 h. Following this, the synthetic brine was drained off and the process was repeated at least twice to remove microorganisms associated with the aqueous phase. Oil-associated biomass was dislodged from the oil by centrifuging at 25,000×g for 20 min in 250 ml centrifuge bottles containing 50–100 ml of sterile synthetic brine (depending on the amount of oil). The oil-containing supernatant was removed, 25–30 ml of sterile synthetic brine was added to the pellet and the tube was agitated for 60 s with a Vortex mixer to suspend the cells. The cell suspension was transferred to 50 ml centrifuge tubes and re-centrifuged. Sterile synthetic brine (1 ml) was added and the samples were processed by microcentrifugation as indicated above. The entire process was replicated five times, so that totally ten samples of biomass were collected (five samples of oil-associated biomass and five samples of water-associated biomass) to allow analysis of the composition of five OCs and five ACs (Table 1).

Table 1 The numbers of reads identified at the phylum (NP), class (NC) and genus (NG) levels
Full size table

DNA isolation

DNA was isolated according to Marmur (1961), with a bead beating step modification. Oil- and water-associated biomass was re-suspended in 280 μl 0.15 M NaCl and 0.1 M EDTA (pH 8) and subjected to digestion with lysozyme (20 μl of 5 mg/ml, at 37°C for 10 min or longer). Samples were then transferred into 2 ml bead-containing lysing matrix tubes (MP Biomedicals) and subjected to bead-beating in a FastPrep-24 homogenizer (MP Biomedicals) for 1 min; subsequently, they were centrifuged at 17,000×g for 1 min and transferred to microcentrifuge tubes; 72 μl 5 M NaClO4 and 420 μl chloroform-isoamyl alcohol (24:1) were added and the mix was placed on a rotating wheel for a minimum of 1 h. The mix was centrifuged for 3.5 min at 17,000×g in the same tube at 20°C; 350 μl of the top layer was transferred to another microcentrifuge tube and 700 μl of 95% ethanol was added. Following centrifugation (15 min, 17,000×g, 4°C) the pellet was dissolved in 200 μl of TE (10 mM Tris, 0.1 mM EDTA pH 8) and incubated at 20°C with DNAse-free RNAse (25 μl, 50 μg/ml) and proteinase K (10 μl, 20 mg/ml) sequentially, with each for 1 h. Following phenol extraction and ethanol precipitation, the air-dried DNA pellets were dissolved in 30–50 μl of TE-buffer, depending on the size of the pellet.

DNA pyrosequencing

DNA samples were amplified through a two-step PCR amplification. The first PCR (25 cycles) was performed with 16S primers 926Fw16S (AAACTYAAAKGAATTGRCGG) and 1392R16S (ACGGGCGGTGTGTRC). Once the presence of a 500 bp PCR product was confirmed by agarose gel electrophoresis, a second PCR (ten cycles) was performed using FLX titanium primers 454T_RA_X and 454T_FwB with the sequences for 926Fw16S and 1392R16S as their 3′-ends. Primer 454T_RA_X had a 25 nucleotide A-adaptor (CGTATCGCCTCCCTCGCGCCATCAG) and a ten-nucleotide multiplex identifier barcode sequence. Primer 454T_FwB had a 25 nucleotide B-adaptor sequence (CTATGCGCCTTGCCAGCCCGCTCAG). The PCR product was checked by agarose gel electrophoresis, purified with QIAquick PCR Purification Kit (Qiagen), and the concentration of the purified product was determined with Qubit fluorometer (Invitrogen) using Quant-iT™ dsDNA HS Assay Kit (Invitrogen). PCR products (25 μl of 5 ng/μl) were pyrosequenced at the Genome Quebec and McGill University Innovation Centre (Montreal, Quebec) using a Genome Sequencer FLX Instrument with a GS FLX Titanium Series Kit XLR70 (Roche Diagnostics Corporation).

Analysis of pyrosequencing data

Phoenix 2, an in-house developed SSU rRNA gene data analysis pipeline, was used to conduct the analysis. Raw pyrosequence reads were subjected to stringent systematic checks in order to remove low quality reads and minimize sequencing errors that could be introduced during the pyrosequencing process (Huse et al. 2007). Eliminated sequences included ones that: (1) did not perfectly match the adaptor and primer sequences, (2) had ambiguous bases, (3) had an average quality score below 27, (4) contained homopolymer lengths greater than eight, and (5) were shorter than 200 bp after clipping off the primers. The remaining high-quality sequences were compared against the non-redundant SSURef data set of SILVA102 (Pruesse et al. 2007) using the Tera-Blast algorithm on a 16-board TimeLogic Decypher system (Active Motif, Inc.). Sequences that had the best alignment covering less than 90% of the trimmed read length, with greater than 90% sequence identity to the best BLAST match within the matched region and no match to the ends of the sequence, were labeled as potential chimeras and excluded from further analysis. While this might remove some true high-quality BLAST matches, in most cases a low percentage (less than 19.14%) of the sequences were eliminated, indicating that most of the real BLAST hits were included in the final study.

Sequences that passed the quality control and chimerical sequence removal were clustered into OTUs (Operational Taxonomic Units) at 3% distance by using the average linkage algorithm (Schloss and Westcott 2011). After grouping sequences into OTUs, several alpha diversity indices (Table S1), as well as the total numbers of OTUs were calculated for each sample. A taxonomic consensus for each representative sequence from each OTU was derived by using the classifier.otu function implemented in the Mothur software package (Schloss et al. 2009). The entire set of the raw reads is available at the NCBI Sequence Read Archive (SRA) under accession numbers SRR088842, SRR088844, SRR088845, SRR088846, SRR088847, SRR088848, SRR088850, SRR088852, SRR088854 and SRR088855. Table 1 provides the information on the numbers of reads identified at the phylum, class and genus levels for the ten samples that were analyzed.

To explore potential relationships between microbial communities from different environments, samples were clustered into Newick-formatted trees using the UPGMA algorithm with the distance between communities calculated with the thetaYC coefficient as a measurement of similarity between community structures (Yue and Clayton 2005) and the Jaccard index as a measurement of similarity between community memberships in the Mothur software package. The trees were visualized using Dendroscope (Huson et al. 2007b). In addition, parsimony testing (Fitch 1971), as well as weighted unifrac and unweighted unifrac testing (Hamady et al. 2010; Lozupone and Knight 2005) were used to determine the statistical significance of clustering within the trees. Rank abundance plots were built with MEGAN software (Huson et al. 2007a).

Results

Alpha- and beta-diversity

Two distinct OC and AC branch clusters were formed in the J-tree (Fig. 1A). The clustering pattern within the YC-tree (Fig. 1B) was very similar to that in the J-tree. The robustness of the inferred sample relation trees was confirmed using parsimony, weighted unifrac, unweighted unifrac and AMOVA (Analysis of Molecular Variance) hypothesis testing (Table 2). This suggested that the differences between OCs and ACs were statistically significant.

Fig. 1
figure 1

Clustering analysis of pyrosequencing data. A Jaccard sample relation tree built using the UPGMA clustering algorithm based on the classical Jaccard values. B ThetaYC tree built using the UPGMA algorithm based on Yue and Clayton (2005) theta values. Both trees were calculated in Mothur and visualized with Dendroscope software

Full size image
Table 2 Parsimony, weighted unifrac, unweighted unifrac, and AMOVA hypothesis testing results
Full size table

To eliminate possible differences introduced by sampling bias, data for all OC amplicon libraries were pooled together to form the OCP (pooled microbial community associated with the oil phase) and the data for all AC amplicon libraries were pooled together to form the ACP (pooled microbial community associated with the aqueous phase). The alpha-diversity indices (Table S1) showed that the ACP was more diverse than the OCP (e.g. the Shannon-Wiener indices were 4.38 and 3.89, and the Simpson indices were 0.039 and 0.054, respectively). The richness shown by rarefaction curves (Fig. 2) was also higher in the ACP than in the OCP. The non-parametric Kendall tau rank correlation coefficient (Kendall 1938) was calculated by using the R statistical software package (Gentleman 2008). The P-value (0.0022) and tau (0.0551) suggested that the OCP and the ACP were highly independent from each other.

Fig. 2
figure 2

Rarefaction curves for the OCP and the ACP

Full size image

Characterization of the OCP and the ACP

Phylum and class levels

Proteobacteria and Euryarchaeota were found to be the most abundant phyla represented in both the OCP and the ACP (Table 3). The phyla Actinobacteria, Firmicutes and Tenericutes (Table 3: #2, #3 and #1) were more numerous in the OCP, whereas the phyla Spirochaetes and Synergistetes had higher representation in the ACP (Table 3: #10 and #11). At the class level, the Alphaproteobacteria and Methanomicrobia, being the most numerous in both phases, had almost equal representation in the OCP and the ACP (Table 4: #7 and #8). Deltaproteobacteria, Epsilonproteobacteria and Synergistia had consistently higher representation in the ACP, whereas Methanobacteria, Gammaproteobacteria and Clostridia were more numerous in the OCP (Table 4; Figs. S1, S2, S4 and S5).

Table 3 Numbers and fractions of reads for phyla found in the OCP and the ACP
Full size table
Table 4 Numbers and fractions of reads for classes found in the OCP and the ACP
Full size table

Order and genus levels

The most abundant genera identified in the OCP and the ACP are presented in the Fig. S3 and S6. Three genera with the highest combined representation in the OCP and the ACP were all methanogens (Table 5: #27 Methanosaeta, #12 Methanobacterium and #8 Methanolobus). Of these, Methanosaeta was equally represented in two communities, whereas Methanobacterium and Methanolobus had higher representation in the OCP. Two other methanogens of the order Methanosarcinales, #1 Methanosarcina and #6 Methanomethylovorans, were more abundant in the OCP by 77- and 11-fold respectively. In contrast, methanogens of the order Methanomicrobiales, namely genera Methanofollis, Methanocalculus and Methanoculleus, had higher representation in the ACP (Table 5: #30, #44 and #45). Methanospirillum belonging to the order Methanomicrobiales was equally represented in the OCP and the ACP (Table 5: #23). Hence, there appeared to be a distinct distribution of methanogens with the order Methanosarcinales being predominantly oil-attached, except for the acetotrophic methanogen Methanosaeta, and the order Methanomicrobiales being more abundant in the aqueous phase.

Table 5 Numbers and fractions of reads for genera found in the OCP and the ACP
Full size table

Although the class Alphaproteobacteria was equally represented in the OCP and ACP (Table 4), large differences were seen in distributions at the genus level. Representatives of four genera were more abundant in the OCP (Table 5: #2 Sphingomonas, #11 Rhodovulum, #14 Brevundimonas and #15 Stappia), whilst seven genera had higher representation in the ACP (Table 5: #50 Sphingobium, #46 Tistrella, #43 Donghicola, #42 Oceanibaculum, #37 Mesorhizobium, #33 Hyphomonas and #32 Ensifer), and four genera were equally represented in the OCP and the ACP (Table 5: #16 Kaistia, #19 Hyphomicrobium, #22 Rhizobium, and #26 Rhodobacter). Likewise, Betaproteobacteria phylotypes were either oil-associated (Table 5: #4 Azovibrio) or mostly associated with the aqueous phase (#41 Azonexus). Thalassolituus (Table 5: #3) was the most prominent oil-associated representative of the class Gammaproteobacteria. The genera Pseudoxanthomonas and Azomonas (Table 5: #17 and #25) were equally represented in the two communities. Remarkably, all of eight genera identified within the Deltaproteobacteria preferred the aqueous phase (#51 Geobacter, #49 Desulfocapsa, #48 Desulfomicrobium, #47 Desulfobulbus, #38 Syntrophus, #36 Desulfuromonas, #35 Pelobacter and #34 Desulfovibrio). Of these, Geobacter had the least affinity to oil, being 80-fold more abundant in the ACP than in the OCP. The epsilonproteobacterium Arcobacter was also predominantly associated with the aqueous phase (Table 5: #39).

The oil-association of the phyla Firmicutes and Tenericutes (Table 3) was largely because of members of the genus Acetobacterium (Table 5: #7) and Acholeplasma (#9), respectively. Likewise, the preference of the phylum Synergistetes for the aqueous phase was mostly due to members of the genus Thermovirga (Table 5: #40). Among Actinobacteria, Propionicicella was one of the most highly represented genera in the OCP (Table 5: #5); Rhodococcus (Table 5: #10) was also oil-associated; other Actinobacteria identified at the genus level (e.g. #29 Georgenia, Table 5) had about equal representation in the two communities.

Discussion

Why attach to oil?

Oil in the Enermark field, from which the samples used in this study were taken, is produced by water injection. The injected water contains a low concentration of sulfate (1 mM) and nitrate (2 mM). The latter is added to prevent sulfide production by sulfate reducing bacteria or oxidize sulfide that is already present. As indicated elsewhere (Voordouw et al. 2009), both injected electron acceptors are reduced completely along the flow path. As a result, produced waters contain no nitrate or nitrite, no sulfate and on average 0.5 mM sulfide (Voordouw et al. 2009). The predominance of methanogens in the microbial community of produced waters (Table 5) is in agreement with the previously noted depletion of electron acceptors and suggests that oil degradation coupled to methanogenesis is a primary degradative mechanism in the vicinity of producing wells. Enermark oil is heavy, meaning that it is depleted of low molecular weight components such as toluene and enriched in higher molecular weight components such as asphaltenes. One may expect that microorganisms, which act primarily on substrates in the oil phase have an incentive to be oil-attached through hydrophobic cell surfaces. These may be primary oil degraders or utilizers of H2, which has a higher solubility in oil than in water.

Methanogenic oil degradation depends on the activities of: (1) syntrophic bacteria hydrolyzing hydrocarbons to H2, CO2 and acetate, (2) hydrogenotrophic methanogens converting H2 and CO2 to methane, and (3) acetotrophic methanogens converting acetate to methane and CO2 (Zengler et al. 1999; Head et al. 2003; Gieg et al. 2008). Syntrophic hydrocarbon degradation requires effective removal of hydrogen and acetate to drive this thermodynamically uphill reaction (Zengler et al. 1999). The deltaproteobacterium Syntrophus was identified as a primary degrader of a methanogenic hexadecane-degrading consortium (Zengler et al. 1999). However, oil has a wide variety of hydrocarbons including aromatic and polyaromatic hydrocarbons making it likely that other genera are involved as well. For example, Gieg et al. (2008) reported that an oil-utilizing methanogenic consortium consisted of Deltaproteobacteria, Firmicutes, Actinobacteria, Chloroflexi and Bacterioidetes. These phyla were also found in the current study (Table 3).

Oil attachment of H2-utilizers

H2 generated during anaerobic, methanogenic oil degradation or other anaerobic processes will distribute over both oil and aqueous phases, however it is more soluble in oil than in water at 30°C (Cai et al. 2001). An association of obligate hydrogenotrophs, such as hydrogenotrophic methanogens with the oil phase may therefore be expected. Indeed, two of the three most abundant genera identified here were oil-associated hydrogenotrophic methanogens (Table 5: #8 Methanolobus and #12 Methanobacterium). The acetotrophic methanogen Methanosaeta, one of the most abundant genera identified, does not need to be oil-associated, because at the resident pH of 7–8 acetate is more water- than oil-soluble. Methanosarcina species consume both acetate and H2 (Beckmann et al. 2011). However, the strong association of this species with oil (Table 5) indicates that hydrogen may be its preferred substrate in the Enermark field. Species of Methanomethylovorans produce methane from dimethylsulfide, methanethiol and methylamine (Lomans et al. 1999), compounds that dissolve in oil rather than water, explaining the association of this genus with oil (Table 5). The hydrogenotrophic methanogens Methanocalculus and Methanoculleus require acetate as a carbon source but not as a substrate for methanogenesis (Lai et al. 2002, 2004; Mikucki et al. 2003), which may explain their preferential association with the aqueous phase. The hydrogenotrophic acetogen Acetobacterium was also strongly oil-associated (Table 5).

Primary hydrocarbon degraders attached to oil

Gammaproteobacteria

Thalassolituus, the major gammaproteobacterium identified in the Enermark field, was clearly oil-associated. Thalassolituus oleivorans is an obligate oil-degrading bacterium capable of utilizing water-insoluble alkanes (Yakimov et al. 2010). This species was previously found in crude oil-containing marine environments and characterized as an aerobic chemoorganoheterotrophic organism unable to grow by fermentation or nitrate reduction (Yakimov et al. 2010). Oil-attachment of Thalassolituus in the Enermark field indicates that some species of this genus metabolize oil anaerobically.

Alphaproteobacteria

Members of the four oil-associated genera Sphingomonas, Rhodovulum, Brevundimonas, and Stappia have all been shown to degrade hydrocarbons. Sphingomonas spp. are well-known degraders of polyaromatic hydrocarbons. Although generally described as aerobic, it has been previously identified in anaerobic oil-degrading consortia (Zhang et al. 2011). Oil-degrading species of Rhodovulum were isolated recently by Teramoto et al. (2010), whereas species of Brevundimonas and Stappia have also been characterized as oil-degraders (Al-Awadhi et al. 2007; Chaîneau et al. 1999).

Betaproteobacteria

Azovibrio, found in this study to be associated with the oil phase, is closely related to Azoarcus (and also related to Thauera); some species of Azoarcus are known to degrade aromatic hydrocarbons under denitrifying conditions (Reinhold-Hurek and Hurek 2006), suggesting that Azovibrio may possibly also be able to degrade oil.

Actinobacteria

Some representatives of this phylum/class (for example, some species of Rhodococcus) have been characterized as degraders of poorly soluble in water organic compounds (Andreoni et al. 2000) and our finding that most Actinobacteria present in the Enermark field are oil-associated corresponds well to this information. Propionicicella was the most numerous genus of Actinobacteria in the OCP, but it was not abundant in the ACP. Interestingly, a close relative of this genus, Propionicimonas, was found in the production water from another Canadian oil reservoir (Grabowski et al. 2005a); its known species, Propionicimonas paludicola, has been described as a fermenting bacterium (Akasaka et al. 2003), whose presence in the oil reservoir production water/oil mixture was not easy to explain (Grabowski et al. 2005a). Propionicicella superfundia was found to be a chlorosolvent-tolerant, propionate-forming, facultative anaerobic bacterium (Bae et al. 2006). Unable to directly biotransform chlorinated solvents, it nevertheless has been thought to play a role in their biodegradation (Bae et al. 2006). It is not likely that chlorosolvents have been present in the produced water–oil mixture. However the apparent hydrophobicity of the cells of this bacterium may assist it in consuming hydrocarbons that are insoluble in water.

Tenericutes

Acholeplasma was one of the ten most abundant genera represented in the OCP. Representatives of this genus are known to be carbohydrate fermenters (Green et al. 2008). Somewhat unexpected in an oil field, Acholeplasma was previously found in a similar environment, a coal-bed methane well (Green et al. 2008). Its association with the oil phase suggests that species inhabiting this oil field may be able to metabolize hydrocarbons.

Microorganisms associated with the aqueous phase

Deltaproteobacteria

All eight identified genera of the Deltaproteobacteria (Table 5) were associated with the aqueous phase. This is understandable for those using organic acids such as acetate as electron donor for sulfate or Fe(III) reduction (Geobacter, Pelobacter, Desulfobulbus, Desulfomicrobium, Desulfuromonas, Desulfovibrio and Desulfocapsa). One might have expected the genus Syntrophus, implicated as the primary degrader in anaerobic hexane degradation (Zengler et al. 1999), to be oil-associated; however this hydrophilic bacterium may be able to access water-insoluble substrates through a biosurfactant-mediated oil pseudosolubilization (2008).

Alphaproteobacteria

Of the seven genera with representatives which preferred the aqueous phase (Sphingobium, Tistrella, Donghicola, Oceanibaculum, Mesorhizobium, Hyphomonas and Ensifer), some have been described as hydrocarbon degraders. Sphingobium species were reported to consume both water-insoluble aromatic hydrocarbons and such water-soluble substances as benzoate (Liang and Lloyd-Jones 2010). Being associated with the aqueous phase suggests that Sphingobium strains may preferentially consume water-soluble organics in the Enermark oil field. Tistrella increased the speed of phenanthrene degradation by another bacterium, Sphingomonas; however by itself it could not degrade this compound (Zhao et al. 2008) and the results of our study suggest that its primary metabolic substrate(s) must be water-soluble; alternatively, it may be able to produce biosurfactants for accessing water-insoluble organics. We came to the same conclusion regarding Oceanibaculum, some species of which have been characterized by Lai et al. (2009) and Dong et al. (2010). Donghicola has been thought to play a role in oil degradation, however it could also use water-soluble acetate (Tan et al. 2009). To our knowledge, Mesorhizobium, Hyphomonas and Ensifer have not been reported to degrade hydrocarbons.

Bacterioidetes and Synergistetes

Petrimonas was found in our study to have representation in the ACP about twice as large as in the OCP (Table 5: #31). The species Petrimonas sulfuriphila was shown by Grabowski et al. (2005b) to consume acetate and other organic acids. The phylum Bacteroidetes, to which Petrimonas belongs, was about one and half times more abundant in the ACP than the OCP (Table 3), showing that it is represented in the Enermark oil field largely by hydrophilic bacteria whose major metabolic substrates are, most likely, organic acids.

Thermovirga lienii was isolated by Dahle and Birkeland (2006) from a North Sea oil well and shown to consume some organic acids, and this fact corresponds well to our finding that Thermovirga is a bacterium associated with the aqueous phase.

Role of diazotrophs

Members of the order Rhizobiales, generally known to be diazotrophs, were found either to be nearly equally represented in the OC and the AC (Table 5: #16 Kaistia #19 Hyphomicrobium and #22 Rhizobium, the latter genus being one of the most represented bacteria in both the OC and the AC) or associated with the aqueous phase (Table 5: #32 Ensifer and #37 Mesorhizobium). Although ammonium was found in the Enermark field in submillimolar concentrations, preferential denitrification of the injected nitrate combined with anammox (Cornish-Shartau et al. 2010) can lead to a lack of NH3 and abundance of N2 in some zones of the oil reservoir. Therefore, Rhizobiales, together with some other bacteria (e.g. Azovibrio and Azonexus) may indeed convert N2 produced by nitrate-reducing bacteria to NH3 for biomass formation.

Concluding remarks

We have shown that a distinct microbial community is attached to oil. Many primary hydrocarbon degraders, especially among the Gamma- and Alphaproteobacteria, were found to be biased towards oil attachment. Hydrogenotrophs also appeared to attach to oil, suggesting that the oil phase is a source of hydrogen. Although oil degradation in the Enermark field is anaerobic, we would like to note with respect to aerobic hydrocarbon degradation that O2, like hydrogen, is more soluble in oil than in water, providing an additional incentive for some aerobic hydrocarbon degraders to associate with oil.

Concerning future research, it is important to note that the composition of oil has been reported to influence microbial adhesion to it. In particular, the presence of asphaltenes and resins may decrease affinity of microbial cells to hydrocarbons, whereas the presence of toluene may lead to its increase (Zoueki et al. 2010). Therefore, analysis of the oil composition in future experiments may help explain possible differences in the compositions of the OC and the AC from different oil fields.