Marine Pollution Bulletin 90 (2015) 106–114
Contents lists available at ScienceDirect
Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul
Initial community and environment determine the response of bacterial communities to dispersant and oil contamination Alice C. Ortmann a,b,⇑, YueHan Lu c a
Department of Marine Sciences, University of South Alabama, 307 University Blvd, Mobile, AL 36688, United States Dauphin Island Sea Lab, 101B Bienville Blvd, Dauphin Island, AL 36528, United States c Department of Geological Sciences, University of Alabama, Tuscaloosa, AL 35487, United States b
a r t i c l e
i n f o
Article history: Available online 6 December 2014 Keywords: Deepwater Horizon Bacterial community Oil spill 16S rRNA Gulf of Mexico Particulate alkanes
a b s t r a c t Bioremediation of seawater by natural bacterial communities is one potential response to coastal oil spills, but the success of the approach may vary, depending on geographical location, oil composition and the timing of spill. The short term response of coastal bacteria to dispersant, oil and dispersed oil was characterized using 16S rRNA gene tags in two mesocosm experiments conducted two months apart. Despite differences in the amount of oil-derived alkanes across the treatments and experiments, increases in the contributions of hydrocarbon degrading taxa and decreases in common estuarine bacteria were observed in response to dispersant and/or oil. Between the two experiments, the direction and rates of changes in particulate alkane concentrations differed, as did the magnitude of the bacterial response to oil and/or dispersant. Together, our data underscore large variability in bacterial responses to hydrocarbon pollutants, implying that bioremediation success varies with starting biological and environmental conditions. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Following the release of oil from the Deepwater Horizon in 2010, strong responses in the microbial community were detected in the deep-water bacterial community in the northern Gulf of Mexico. The community diversity changed rapidly following the input of oil from the leaking well and the addition of dispersants (Hazen et al., 2010; Valentine et al., 2010; Kessler et al., 2011; Redmond and Valentine, 2011). The rapid response was likely due to a long history of natural seeps in the region (MacDonald et al., 2002), which likely led to the development of a bacterial community with high hydrocarbon-oxidizing potential (Haritash and Kaushik, 2009). The deep-water plume was dominated by a few types of Bacteria, with some samples dominated by a single operational taxonomic unit (OTU) representing up to 90% of the total sequences. Outside the plume, this sequence contributed only 5% of the total sequences (Hazen et al., 2010; Mason et al., 2012). This OTU was identified as a member of the Oceanospirillales, and sequencing of single cells indicated the potential for these organisms to degrade n-alkanes and cycloalkanes. Several inor⇑ Corresponding author at: Dauphin Island Sea Lab, 101 Bienville Blvd, Dauphin Island, AL 36528, United States. Tel.: +1 251 861 2141x7526. E-mail address: [email protected] (A.C. Ortmann). http://dx.doi.org/10.1016/j.marpolbul.2014.11.013 0025-326X/Ó 2014 Elsevier Ltd. All rights reserved.
ganic nutrient transporters were also identified, suggesting a requirement for ammonium and phosphate (Mason et al., 2012). Other members of the deep-water plume included Cycloclasticus spp., which are known to degrade aromatic hydrocarbons (Head et al., 2006; Teira et al., 2007), and Colwellia spp., which were likely metabolizing ethane and propane (Redmond and Valentine, 2011). Both Cycloclasticus spp. and Colwellia spp. appeared to increase as the distance from the wellhead increased, following the bloom of the Oceanospirillales. In the older plume, furthest from the wellhead, an increase in methanotrophs was detected, suggesting a slower response of this community to the inputs of methane gas (Kessler et al., 2011). In surface waters near the well site, a rapid increase in bacterial respiration was observed, but no increase in biomass or cell number occurred (Edwards et al., 2011). The addition of inorganic nutrients to incubations stimulated cell division, suggesting the microbial community in the surface waters was nutrient limited. Observations of phosphate stress supported this conclusion, and several previous studies have indicated the importance of inorganic nutrients in the ability of some Bacteria to respond to oil spills (Jimenez et al., 2007; Jean et al., 2008). The bacterial community in surface waters contaminated with oil and dispersants was determined to be more diverse than the community at depth. Where oil was present as a thin sheen, Gammaproteobacteria, such as
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Alteromonadales and Oceanospirillales, composed about 15% of the community (Redmond and Valentine, 2011). Heavy oiling increased the contribution of Gammaproteobacteria to almost 100%, but Colwellia spp. and Cycloclasticus spp. represented 20, (c) have no ambiguous bases and (d) have homopolymer runs with >4 bases. No tolerance was allowed for mismatches between barcode and forward primers, and reverse primers were removed along with the barcode and forward primer. All sequences from the two experiments were combined into one fasta file for further processing. Sequences were processed using the pick_open_reference_otus.py workflow with the default values and uclust (Edgar, 2010).
108
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Taxonomy was assigned using the August 2013 release of Greengenes (DeSantis et al., 2006) and the naïve Bayesian RDP classifier (Wang et al., 2007). The OTU table was filtered to remove all OTUs represented by a single sequence as well as any OTUs that failed to align with a 16S rRNA database using PyNAST (75% min ID) (Caporaso et al., 2009). The OTU table (30,875 OTUs and 1,371,159 reads) was further filtered to include only OTUs that represented at least 0.001% of the total reads (Bokulich et al., 2013). The final table consisted of 812 OTUs (2.6% of the OTUs) representing 1,134,513 reads (82.7% of the reads). The table was subsampled to an even depth of 2940 sequences per sample and then processed using the summarize_taxa.py script to cluster OTUs based on taxonomic assignments. The family level analysis, representing 75 taxonomic groups, was chosen for all further analysis. Samples from t0 were analyzed to determine if there were significant differences in the starting communities in the two experiments. Estimates of richness (Margalef) and diversity (Shannon diversity) were calculated for each sample and compared using a Students t-test as the data fit the assumptions for parametric analysis. Community structure was compared by calculating the Bray–Curtis similarity metric using the 75 family-level taxa (Primer 6, Primer-E Ltd., UK). Significant differences between experiments were identified using ANOSIM (analysis of similarity). A non-parametric t-test calculated using Monte Carlo simulation was carried out in QIIME (group_significance.py) to identify taxa that differed significantly between experiments. P-values were evaluated after a false-discovery rate correction (FDR). Taxa contributing an average of 1% of reads to the total community were identified to illustrate differences in the starting community. The effects of dispersant, oil and dispersed oil on the bacterial community were quantified for each experiment separately. Richness and diversity estimates were calculated for each sample collected at t0, t24 and t72. As the data met the assumptions for parametric tests, significant effects of treatment on these metrics were detected using a Repeated Measures ANOVA in JMP 9.0 (SAS Institute Inc., USA), a = 0.05. Community structure was analyzed using the Bray–Curtis similarity metric calculated using the 75 family-level taxa (Primer 6). Significant differences between treatments and samples from t24 and t72 were detected using a 2-way ANOSIM using treatment and time as factors. Pairwise tests for differences in treatment were carried out when the treatment R was significant (R > 0.3, p < 0.05). To identify taxa that had the largest response to time or treatment, the percent change in community composition relative to t0 was calculated (% contribution at t24 or t72 for each treatment % contribution at t0). Taxa that increased or decreased by at least 5% (a change of at least 147 reads) were identified and patterns across time and treatment were determined. Sequences were submitted to the Sequence Read Archive (SRA) at NCBI under the study accession number PRJNA188783 (Runs: SRR833379-SRR833484). 2.4. Particulate hydrocarbon collection and analysis Two out of the 5 mesocosms for each treatment were sampled for particulate hydrocarbon analysis. All sampling and analysis apparatus to be in direct contact with samples were either combusted at 450 °C for 5 h (glass materials), or acid soaked (10% HCl) and thoroughly rinsed with Milli-Q water (plastic materials). At day 0 (t0) and day 10 (June) or day 7 (August) (tn), 2 L of water from ca. 10 cm below the surface was filtered through a 142 mm, 0.7 lm glass microfiber filter (Whatman GF/F). Filters were folded, placed in pre-combusted aluminum foil pouches, and put on ice for transportation. Samples were frozen within 6 h of collection. Hydrocarbon analysis followed the method described by Risdon et al. (2008) with modifications. The filters were shredded with
solvent-washed forceps and mixed with Hydromatrix (Agilent Technologies) to remove water. Normal hexadecane-d34 (n-C16D34) was added as a surrogate prior to the extraction to standardize the extraction efficiency. Samples were then ultrasonically extracted with 4 ml acetone for 2 min at 20 °C to ensure that the surrogate was thoroughly mixed into the samples. Twenty ml of acetone:hexane (1:1, vol:vol) was then added to the samples, followed by two ultrasonic extractions for 10 min at 20 °C. Solvent-washed copper turnings were added to the extracts to remove sulfur. After storage at 4 °C overnight, the extracts were concentrated with gentle ultra-high purity (UHP) nitrogen stream, followed by addition of 3 ml acetone, 5 ml hexane, 4 ml Milli-Q water and a spatula of sodium chloride. The mixtures were manually shaken for 30 s and allowed to settle for 20 min, with the top hexane layer siphoned as the hydrocarbon fraction. The hydrocarbon fraction was then separated by silica column chromatography into aliphatic and aromatic hydrocarbon fractions, which were eluted by 10 ml of hexane and 12 ml of dichloromethane, respectively. Aliphatic fractions were quantified and identified using a Shimadzu GC-2014 gas chromatograph (GC), and compound identification was confirmed using an Agilent 6890 GC-5973MSD mass spectrometer. The GC oven was set at 50 °C, held for 1 min, and then increased by 6 °C min 1 to 310 °C and held for 15 min. Helium was used as the carrier gas at a flow rate of 1.48 ml min 1. The quantification of objective compounds was done by comparing their peak areas to the peak areas of n-C16D34, with the correction of instrumental responses of various compounds that were obtained by regularly running a mixture containing a C7–C40 saturated alkane standard mixture (Sigma–Aldrich) and n-C16D34. As the data were not normally distributed, comparisons of measured parameters across the treatments or different times points were conducted using a non-parametric Kruskal–Wallis test (for P3 groups) or Mann–Whitney U test (for 2 groups). The significance level, a, was set at 0.05. 3. Results 3.1. Initial community diversity Two estimates of diversity, richness (Margalef) and community diversity (Shannon diversity), were calculated for the initial bacterial communities in June and August (Table 1). There was no significant difference in either of the metrics between the two experiments (Students t-test, p > 0.05). An ANOSIM analysis of the t0 communities using the Bray Curtis similarity metric indicated that the community structures were significantly different between June and August experiments (ANOSIM R = 0.684, p = 0.008). Comparison of taxa in the two communities using a nonparametric t-test (Monte Carlo simulation) and the Benjamini–Hochberg FDR procedure for multiple comparisons identified only three taxa with significantly different abundances between June and August (Supplemental Table 1). Two of these, the unclassified Betaproteobacteria and the C111 group of the Acidimicrobiales, averaged 0.05). Errors are standard deviations of the five samples. Experiment
Margalef richness
Shannon diversity
June August
5.33 ± 0.55 5.53 ± 0.57
2.21 ± 0.26 2.28 ± 0.16
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Verrucomicrobia, averaged 2.5% of the reads in the ten t0 samples, and was significantly higher in June than in August. Between June and August, 11 taxa were identified that contributed an average of at least 1.0% of the total reads at t = 0. These taxa, with the exception of the Puniceicoccaceae, did not have significantly different percent contributions between the two experiments (Fig. 1). Although only three taxa showed significant differences between experiments, small differences in the abundances of a large number of taxa likely drove the differences in community structure detected by ANOSIM.
Synechococcaceae
August (Fig. 3). In June the Pelagibacteraceae increased slightly in the controls, while a small decrease in August was observed in the control treatments. Rhodobacteraceae also decreased in the dispersant, oil and dispersed oil treatments in both experiments; however, the largest decreases were actually in the control treatments. At the same time, the percentage of unclassified Gammaproteobacteria increased strongly in the dispersant, oil and dispersed oil treatments, with little or no change in the contribution of this group in the controls. The Oceanospirillaceae increased in June in the presence of oil, with small increases in the dispersant and dispersed oil treatments. In August, this family showed a larger response, increasing in the presence of dispersant, oil and dispersed oil. Interestingly, the increase in the Oceanospirillaceae in the dispersant only treatment at t24 was followed by a sharp decrease to levels observed at the beginning of the experiment by t72. The fifth taxon that changed by at least 5% over the course of both experiments was the unclassified Bacteria group. While this taxon generally increased in percent contribution in June, it tended to decrease in August. As the identities of the organisms that fall in the group are unknown, it is likely that different species with different metabolic capabilities were present in the two experiments. In June, two other taxa were found to increase or decrease by more than 5% during the experiment. The percentage of Alteromonadaceae increased in the dispersed oil relative to t0, but no increase was observed in the dispersant or oil only treatments. In August, the Alteromonadaceae increased slightly in the dispersant and dispersed oil treatments, but the biggest change was only 2.4%. The Puniceicoccaceae were also found to change by >5% during the June experiment. This family represented 4.8% of the starting community in June, but only 0.3% of the community in August. In the June experiment, Puniceicoccaceae increased in the control and oil treatments by t24, but decreased to starting abundances or less by t72. Taxa observed to change by >5% in August, but not June, included the unclassified Alphaproteobacteria, unclassified Proteobacteria, the Halomonadaceae and the Methylophilaceae. The first two taxa were observed to increase in the dispersant and dispersed oil treatments at t72 in August. Slight increases in these two taxa also occurred in dispersant and dispersed oil treatments in June, but the increases were 0.8. To identify the taxa that had the largest responses to treatments or time, the change in the percent contribution to the community relative to t0 was calculated for each taxon (Supplementary Tables 2 and 3). Taxa increasing or decreasing by an average of at least 5% in one of the 8 time/treatment categories were identified. In June, seven taxa were identified that fit this criteria (Fig. 3A), while nine taxa were identified from the August experiment (Fig. 3B). Five of these taxa were identified in both experiments. Compared to the t0 samples, the contribution of Pelagibacteraceae decreased in the presence of oil, dispersant and dispersed oil in both June and
Puniceicoccaceae
Average % contribution to t 0 community
100 Flavobacteriaceae 90 80 70 60
Unclassified Gammaproteobacteria
50
Methylophilaceae
40
Pelagibacteraceae
30
Rhodobacteraceae
20
Unclassified Alphaproteobacteria
10
Unclassified Proteobacteria
0 June
August
109
Unclassified Bacteria
Fig. 1. Average contribution of 11 taxa that were represented by at least 1% of the total reads in the 10 samples collected at t0. These taxa represent 94% of the reads in June and 90% of the reads in August.
Alkane compounds detected in samples included both normal and isoprenoid compounds, and the carbon number ranged between 10 and 33. Comparison of the four treatments indicated that the dispersed oil mesocosms had the highest concentrations of total particulate alkanes at t0 (Fig. 4), averaging 1.90 107 ± 8.25 106 ng L 1, which was 450 times greater than in the oil and dispersant mesocosms and ca. 4000 times higher than the control mesocosms. This pattern shows that the dispersant facilitated rapid delivery of oil into the water column. Over time, the concentrations of total particulate alkanes changed depending on the treatment. Only the dispersant and the dispersed oil treatments showed apparent decreases in total particulate alkanes (Fig. 4), with a 74% ± 18% decrease in the
110
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Margalef Richness
June 8.0
B
7.0 6.0
A
A
ABB
B
A
AB B
B
5.0 4.0 3.0 t=0
Shannon Diversity
August
A
3.0
t=24
t=0
t=24
t=72
t=0
t=24
t=72
t=72 Control
C
D
Dispersant
2.5 Oil
2.0
Dispersed Oil
1.5 t=0
t=24
t=72
Fig. 2. Mean community diversity estimates for samples collected from all treatments. Letters indicate significant differences among estimates based on Repeated Measures ANOVA. (A and B) Margalef richness (C and D) Shannon diversity.
Table 2 ANOSIM R values from a 2-way analysis comparing the effects of treatment and time on community composition for samples collected at t24 and t72. All p = 0.001 except where indicated.
Time Treatment Control vs. dispersant Control vs. oil Control vs. dispersed oil Dispersant vs. oil Dispersant vs. dispersed oil Oil vs. dispersed oil
June
August
0.872 0.772 0.904 0.365 (p = 0.009) 0.994 0.875 0.632 1.000
0.653 0.841 0.978 0.889 1.000 0.816 0.497 (p = 0.002) 0.891
experiments, the rate of changes in particulate alkanes differed (Fig. 5). Similar patterns, however, were found in compound distribution variation across the three treatments. The ratio of shortchain to long-chain alkanes (
Contents lists available at ScienceDirect
Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul
Initial community and environment determine the response of bacterial communities to dispersant and oil contamination Alice C. Ortmann a,b,⇑, YueHan Lu c a
Department of Marine Sciences, University of South Alabama, 307 University Blvd, Mobile, AL 36688, United States Dauphin Island Sea Lab, 101B Bienville Blvd, Dauphin Island, AL 36528, United States c Department of Geological Sciences, University of Alabama, Tuscaloosa, AL 35487, United States b
a r t i c l e
i n f o
Article history: Available online 6 December 2014 Keywords: Deepwater Horizon Bacterial community Oil spill 16S rRNA Gulf of Mexico Particulate alkanes
a b s t r a c t Bioremediation of seawater by natural bacterial communities is one potential response to coastal oil spills, but the success of the approach may vary, depending on geographical location, oil composition and the timing of spill. The short term response of coastal bacteria to dispersant, oil and dispersed oil was characterized using 16S rRNA gene tags in two mesocosm experiments conducted two months apart. Despite differences in the amount of oil-derived alkanes across the treatments and experiments, increases in the contributions of hydrocarbon degrading taxa and decreases in common estuarine bacteria were observed in response to dispersant and/or oil. Between the two experiments, the direction and rates of changes in particulate alkane concentrations differed, as did the magnitude of the bacterial response to oil and/or dispersant. Together, our data underscore large variability in bacterial responses to hydrocarbon pollutants, implying that bioremediation success varies with starting biological and environmental conditions. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Following the release of oil from the Deepwater Horizon in 2010, strong responses in the microbial community were detected in the deep-water bacterial community in the northern Gulf of Mexico. The community diversity changed rapidly following the input of oil from the leaking well and the addition of dispersants (Hazen et al., 2010; Valentine et al., 2010; Kessler et al., 2011; Redmond and Valentine, 2011). The rapid response was likely due to a long history of natural seeps in the region (MacDonald et al., 2002), which likely led to the development of a bacterial community with high hydrocarbon-oxidizing potential (Haritash and Kaushik, 2009). The deep-water plume was dominated by a few types of Bacteria, with some samples dominated by a single operational taxonomic unit (OTU) representing up to 90% of the total sequences. Outside the plume, this sequence contributed only 5% of the total sequences (Hazen et al., 2010; Mason et al., 2012). This OTU was identified as a member of the Oceanospirillales, and sequencing of single cells indicated the potential for these organisms to degrade n-alkanes and cycloalkanes. Several inor⇑ Corresponding author at: Dauphin Island Sea Lab, 101 Bienville Blvd, Dauphin Island, AL 36528, United States. Tel.: +1 251 861 2141x7526. E-mail address: [email protected] (A.C. Ortmann). http://dx.doi.org/10.1016/j.marpolbul.2014.11.013 0025-326X/Ó 2014 Elsevier Ltd. All rights reserved.
ganic nutrient transporters were also identified, suggesting a requirement for ammonium and phosphate (Mason et al., 2012). Other members of the deep-water plume included Cycloclasticus spp., which are known to degrade aromatic hydrocarbons (Head et al., 2006; Teira et al., 2007), and Colwellia spp., which were likely metabolizing ethane and propane (Redmond and Valentine, 2011). Both Cycloclasticus spp. and Colwellia spp. appeared to increase as the distance from the wellhead increased, following the bloom of the Oceanospirillales. In the older plume, furthest from the wellhead, an increase in methanotrophs was detected, suggesting a slower response of this community to the inputs of methane gas (Kessler et al., 2011). In surface waters near the well site, a rapid increase in bacterial respiration was observed, but no increase in biomass or cell number occurred (Edwards et al., 2011). The addition of inorganic nutrients to incubations stimulated cell division, suggesting the microbial community in the surface waters was nutrient limited. Observations of phosphate stress supported this conclusion, and several previous studies have indicated the importance of inorganic nutrients in the ability of some Bacteria to respond to oil spills (Jimenez et al., 2007; Jean et al., 2008). The bacterial community in surface waters contaminated with oil and dispersants was determined to be more diverse than the community at depth. Where oil was present as a thin sheen, Gammaproteobacteria, such as
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Alteromonadales and Oceanospirillales, composed about 15% of the community (Redmond and Valentine, 2011). Heavy oiling increased the contribution of Gammaproteobacteria to almost 100%, but Colwellia spp. and Cycloclasticus spp. represented 20, (c) have no ambiguous bases and (d) have homopolymer runs with >4 bases. No tolerance was allowed for mismatches between barcode and forward primers, and reverse primers were removed along with the barcode and forward primer. All sequences from the two experiments were combined into one fasta file for further processing. Sequences were processed using the pick_open_reference_otus.py workflow with the default values and uclust (Edgar, 2010).
108
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Taxonomy was assigned using the August 2013 release of Greengenes (DeSantis et al., 2006) and the naïve Bayesian RDP classifier (Wang et al., 2007). The OTU table was filtered to remove all OTUs represented by a single sequence as well as any OTUs that failed to align with a 16S rRNA database using PyNAST (75% min ID) (Caporaso et al., 2009). The OTU table (30,875 OTUs and 1,371,159 reads) was further filtered to include only OTUs that represented at least 0.001% of the total reads (Bokulich et al., 2013). The final table consisted of 812 OTUs (2.6% of the OTUs) representing 1,134,513 reads (82.7% of the reads). The table was subsampled to an even depth of 2940 sequences per sample and then processed using the summarize_taxa.py script to cluster OTUs based on taxonomic assignments. The family level analysis, representing 75 taxonomic groups, was chosen for all further analysis. Samples from t0 were analyzed to determine if there were significant differences in the starting communities in the two experiments. Estimates of richness (Margalef) and diversity (Shannon diversity) were calculated for each sample and compared using a Students t-test as the data fit the assumptions for parametric analysis. Community structure was compared by calculating the Bray–Curtis similarity metric using the 75 family-level taxa (Primer 6, Primer-E Ltd., UK). Significant differences between experiments were identified using ANOSIM (analysis of similarity). A non-parametric t-test calculated using Monte Carlo simulation was carried out in QIIME (group_significance.py) to identify taxa that differed significantly between experiments. P-values were evaluated after a false-discovery rate correction (FDR). Taxa contributing an average of 1% of reads to the total community were identified to illustrate differences in the starting community. The effects of dispersant, oil and dispersed oil on the bacterial community were quantified for each experiment separately. Richness and diversity estimates were calculated for each sample collected at t0, t24 and t72. As the data met the assumptions for parametric tests, significant effects of treatment on these metrics were detected using a Repeated Measures ANOVA in JMP 9.0 (SAS Institute Inc., USA), a = 0.05. Community structure was analyzed using the Bray–Curtis similarity metric calculated using the 75 family-level taxa (Primer 6). Significant differences between treatments and samples from t24 and t72 were detected using a 2-way ANOSIM using treatment and time as factors. Pairwise tests for differences in treatment were carried out when the treatment R was significant (R > 0.3, p < 0.05). To identify taxa that had the largest response to time or treatment, the percent change in community composition relative to t0 was calculated (% contribution at t24 or t72 for each treatment % contribution at t0). Taxa that increased or decreased by at least 5% (a change of at least 147 reads) were identified and patterns across time and treatment were determined. Sequences were submitted to the Sequence Read Archive (SRA) at NCBI under the study accession number PRJNA188783 (Runs: SRR833379-SRR833484). 2.4. Particulate hydrocarbon collection and analysis Two out of the 5 mesocosms for each treatment were sampled for particulate hydrocarbon analysis. All sampling and analysis apparatus to be in direct contact with samples were either combusted at 450 °C for 5 h (glass materials), or acid soaked (10% HCl) and thoroughly rinsed with Milli-Q water (plastic materials). At day 0 (t0) and day 10 (June) or day 7 (August) (tn), 2 L of water from ca. 10 cm below the surface was filtered through a 142 mm, 0.7 lm glass microfiber filter (Whatman GF/F). Filters were folded, placed in pre-combusted aluminum foil pouches, and put on ice for transportation. Samples were frozen within 6 h of collection. Hydrocarbon analysis followed the method described by Risdon et al. (2008) with modifications. The filters were shredded with
solvent-washed forceps and mixed with Hydromatrix (Agilent Technologies) to remove water. Normal hexadecane-d34 (n-C16D34) was added as a surrogate prior to the extraction to standardize the extraction efficiency. Samples were then ultrasonically extracted with 4 ml acetone for 2 min at 20 °C to ensure that the surrogate was thoroughly mixed into the samples. Twenty ml of acetone:hexane (1:1, vol:vol) was then added to the samples, followed by two ultrasonic extractions for 10 min at 20 °C. Solvent-washed copper turnings were added to the extracts to remove sulfur. After storage at 4 °C overnight, the extracts were concentrated with gentle ultra-high purity (UHP) nitrogen stream, followed by addition of 3 ml acetone, 5 ml hexane, 4 ml Milli-Q water and a spatula of sodium chloride. The mixtures were manually shaken for 30 s and allowed to settle for 20 min, with the top hexane layer siphoned as the hydrocarbon fraction. The hydrocarbon fraction was then separated by silica column chromatography into aliphatic and aromatic hydrocarbon fractions, which were eluted by 10 ml of hexane and 12 ml of dichloromethane, respectively. Aliphatic fractions were quantified and identified using a Shimadzu GC-2014 gas chromatograph (GC), and compound identification was confirmed using an Agilent 6890 GC-5973MSD mass spectrometer. The GC oven was set at 50 °C, held for 1 min, and then increased by 6 °C min 1 to 310 °C and held for 15 min. Helium was used as the carrier gas at a flow rate of 1.48 ml min 1. The quantification of objective compounds was done by comparing their peak areas to the peak areas of n-C16D34, with the correction of instrumental responses of various compounds that were obtained by regularly running a mixture containing a C7–C40 saturated alkane standard mixture (Sigma–Aldrich) and n-C16D34. As the data were not normally distributed, comparisons of measured parameters across the treatments or different times points were conducted using a non-parametric Kruskal–Wallis test (for P3 groups) or Mann–Whitney U test (for 2 groups). The significance level, a, was set at 0.05. 3. Results 3.1. Initial community diversity Two estimates of diversity, richness (Margalef) and community diversity (Shannon diversity), were calculated for the initial bacterial communities in June and August (Table 1). There was no significant difference in either of the metrics between the two experiments (Students t-test, p > 0.05). An ANOSIM analysis of the t0 communities using the Bray Curtis similarity metric indicated that the community structures were significantly different between June and August experiments (ANOSIM R = 0.684, p = 0.008). Comparison of taxa in the two communities using a nonparametric t-test (Monte Carlo simulation) and the Benjamini–Hochberg FDR procedure for multiple comparisons identified only three taxa with significantly different abundances between June and August (Supplemental Table 1). Two of these, the unclassified Betaproteobacteria and the C111 group of the Acidimicrobiales, averaged 0.05). Errors are standard deviations of the five samples. Experiment
Margalef richness
Shannon diversity
June August
5.33 ± 0.55 5.53 ± 0.57
2.21 ± 0.26 2.28 ± 0.16
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Verrucomicrobia, averaged 2.5% of the reads in the ten t0 samples, and was significantly higher in June than in August. Between June and August, 11 taxa were identified that contributed an average of at least 1.0% of the total reads at t = 0. These taxa, with the exception of the Puniceicoccaceae, did not have significantly different percent contributions between the two experiments (Fig. 1). Although only three taxa showed significant differences between experiments, small differences in the abundances of a large number of taxa likely drove the differences in community structure detected by ANOSIM.
Synechococcaceae
August (Fig. 3). In June the Pelagibacteraceae increased slightly in the controls, while a small decrease in August was observed in the control treatments. Rhodobacteraceae also decreased in the dispersant, oil and dispersed oil treatments in both experiments; however, the largest decreases were actually in the control treatments. At the same time, the percentage of unclassified Gammaproteobacteria increased strongly in the dispersant, oil and dispersed oil treatments, with little or no change in the contribution of this group in the controls. The Oceanospirillaceae increased in June in the presence of oil, with small increases in the dispersant and dispersed oil treatments. In August, this family showed a larger response, increasing in the presence of dispersant, oil and dispersed oil. Interestingly, the increase in the Oceanospirillaceae in the dispersant only treatment at t24 was followed by a sharp decrease to levels observed at the beginning of the experiment by t72. The fifth taxon that changed by at least 5% over the course of both experiments was the unclassified Bacteria group. While this taxon generally increased in percent contribution in June, it tended to decrease in August. As the identities of the organisms that fall in the group are unknown, it is likely that different species with different metabolic capabilities were present in the two experiments. In June, two other taxa were found to increase or decrease by more than 5% during the experiment. The percentage of Alteromonadaceae increased in the dispersed oil relative to t0, but no increase was observed in the dispersant or oil only treatments. In August, the Alteromonadaceae increased slightly in the dispersant and dispersed oil treatments, but the biggest change was only 2.4%. The Puniceicoccaceae were also found to change by >5% during the June experiment. This family represented 4.8% of the starting community in June, but only 0.3% of the community in August. In the June experiment, Puniceicoccaceae increased in the control and oil treatments by t24, but decreased to starting abundances or less by t72. Taxa observed to change by >5% in August, but not June, included the unclassified Alphaproteobacteria, unclassified Proteobacteria, the Halomonadaceae and the Methylophilaceae. The first two taxa were observed to increase in the dispersant and dispersed oil treatments at t72 in August. Slight increases in these two taxa also occurred in dispersant and dispersed oil treatments in June, but the increases were 0.8. To identify the taxa that had the largest responses to treatments or time, the change in the percent contribution to the community relative to t0 was calculated for each taxon (Supplementary Tables 2 and 3). Taxa increasing or decreasing by an average of at least 5% in one of the 8 time/treatment categories were identified. In June, seven taxa were identified that fit this criteria (Fig. 3A), while nine taxa were identified from the August experiment (Fig. 3B). Five of these taxa were identified in both experiments. Compared to the t0 samples, the contribution of Pelagibacteraceae decreased in the presence of oil, dispersant and dispersed oil in both June and
Puniceicoccaceae
Average % contribution to t 0 community
100 Flavobacteriaceae 90 80 70 60
Unclassified Gammaproteobacteria
50
Methylophilaceae
40
Pelagibacteraceae
30
Rhodobacteraceae
20
Unclassified Alphaproteobacteria
10
Unclassified Proteobacteria
0 June
August
109
Unclassified Bacteria
Fig. 1. Average contribution of 11 taxa that were represented by at least 1% of the total reads in the 10 samples collected at t0. These taxa represent 94% of the reads in June and 90% of the reads in August.
Alkane compounds detected in samples included both normal and isoprenoid compounds, and the carbon number ranged between 10 and 33. Comparison of the four treatments indicated that the dispersed oil mesocosms had the highest concentrations of total particulate alkanes at t0 (Fig. 4), averaging 1.90 107 ± 8.25 106 ng L 1, which was 450 times greater than in the oil and dispersant mesocosms and ca. 4000 times higher than the control mesocosms. This pattern shows that the dispersant facilitated rapid delivery of oil into the water column. Over time, the concentrations of total particulate alkanes changed depending on the treatment. Only the dispersant and the dispersed oil treatments showed apparent decreases in total particulate alkanes (Fig. 4), with a 74% ± 18% decrease in the
110
A.C. Ortmann, Y. Lu / Marine Pollution Bulletin 90 (2015) 106–114
Margalef Richness
June 8.0
B
7.0 6.0
A
A
ABB
B
A
AB B
B
5.0 4.0 3.0 t=0
Shannon Diversity
August
A
3.0
t=24
t=0
t=24
t=72
t=0
t=24
t=72
t=72 Control
C
D
Dispersant
2.5 Oil
2.0
Dispersed Oil
1.5 t=0
t=24
t=72
Fig. 2. Mean community diversity estimates for samples collected from all treatments. Letters indicate significant differences among estimates based on Repeated Measures ANOVA. (A and B) Margalef richness (C and D) Shannon diversity.
Table 2 ANOSIM R values from a 2-way analysis comparing the effects of treatment and time on community composition for samples collected at t24 and t72. All p = 0.001 except where indicated.
Time Treatment Control vs. dispersant Control vs. oil Control vs. dispersed oil Dispersant vs. oil Dispersant vs. dispersed oil Oil vs. dispersed oil
June
August
0.872 0.772 0.904 0.365 (p = 0.009) 0.994 0.875 0.632 1.000
0.653 0.841 0.978 0.889 1.000 0.816 0.497 (p = 0.002) 0.891
experiments, the rate of changes in particulate alkanes differed (Fig. 5). Similar patterns, however, were found in compound distribution variation across the three treatments. The ratio of shortchain to long-chain alkanes (
