TY - JOUR
AU - Alessi, Daniel, S
AB - ABSTRACT The response of microbial communities to releases of hydraulic fracturing flowback and produced water (PW) may influence ecosystem functions. However, knowledge of the effects of PW spills on freshwater microbiota is limited. Here, we conducted two separate experiments: 16S rRNA gene sequencing combined with random forests modelling was used to assess freshwater community changes in simulated PW spills by volume from 0.05% to 50%. In a separate experiment, live/dead cell viability in a freshwater community was tested during exposure to 10% PW by volume. Three distinct patterns of microbial community shifts were identified: (i) indigenous freshwater genera remained dominant in <2.5% PW, (ii) from 2.5% to 5% PW, potential PW organic degraders such as Pseudomonas, Rheinheimera and Brevundimonas became dominant, and (iii) no significant change in the relative abundance of taxa was observed in >5% PW. Microbial taxa including less abundant genera such as Cellvibrio were potential bioindicators for the degree of contamination with PW. Additionally, live cells were quickly damaged by adding 10% PW, but cell counts recovered in the following days. Our study shows that the responses of freshwater microbiota vary by spill size, and these responses show promise as effective fingerprints for PW spills in aquatic environments. shale gas, hydraulic fracturing, flowback and produced water spills, environmental microbiota, cell viability, bioindicators INTRODUCTION The use of hydraulic fracturing to extract oil and gas from impermeable shale formations has changed the global energy landscape and secured the energy independence of countries that have historically imported a large fraction of their fuel consumption to meet energy demands (Vidic et al. 2013). Implicit in the rapid expansion of hydraulic fracturing is considerable water use and disposal footprints (Barbot et al. 2013; Vidic et al. 2013; Gagnon et al. 2016). A single shale oil and gas well can consume 13.7–23.8 million litres of freshwater during the fracturing process and subsequently may produce 5–50 million litres of flowback and produced water (PW) (Goss et al. 2015; Kondash and Vengosh 2015; Alessi et al. 2017). Estimates of spill frequencies and volumes vary widely depending on source; for example, the United States Environmental Protection Agency reported the total volume spilled was about 7600 m3 with a median spill of 3.7 m3 between 2005 and 2014 in the USA (U.S. Environmental Protection Agency 2015). In a study by Maloney et al. (2017) of 21 300 unconventional wells in Pennsylvania, Colorado, North Dakota and New Mexico, 6622 reported spills from 2005 to 2014. Another study reported 2–16% of 31 481 shale oil and gas wells in Colorado, New Mexico, North Dakota and Pennsylvania had a spill each year between 2005 and 2014, and the largest spills exceeded 100 m3 (Patterson et al. 2017). Spills are not uncommon and include a high frequency of small incidents along with several large spills per year (Brantley et al. 2014). Given the frequency of PW surface releases to near-surface environments, it is important to understand potential impacts to surface water bodies, soils and aquifers. As shale oil and gas extraction by hydraulic fracturing involves both injected fluids and components indigenous to the target geologic formation, the geochemical composition of PW is often complex, consisting of inorganic elements, petroleum compounds and residual chemical additives (Engle, Cozzarelli and Smith 2014; Akob et al. 2015; Ferrer and Thurman 2015; Flynn et al. 2019). Furthermore, release of PW to surface water bodies and shallow aquifers may cause detrimental effects to aquatic animals and drinking water supplies (Parker et al. 2014; Blewett et al. 2017a,b ; Cozzarelli et al. 2017; Folkerts et al. 2017a,b; Orem et al. 2017; He et al. 2017b, 2018; Hossack et al. 2018; Preston et al. 2019; Smalling et al. 2019; Wang et al. 2019). Understanding the effects of PW spills on the ecosystem microbiota is of environmental and economic importance. Microorganisms are basic units of the food web and drive biogeochemical processes in an ecosystem (Prosser et al. 2007). For example, microbes degrade organic matter, stabilise metals and generate greenhouse gas emissions to the atmosphere (Siddique et al. 2012). Previous studies have shown that fracturing chemicals such as polyethylene glycols (PEGs), isopropanol and petroleum hydrocarbons can be utilised by microorganisms in surface environments at laboratory cultivation conditions (Ulrich et al. 2009; Kekacs et al. 2015; McLaughlin, Borch and Blotevogel 2016). However, PW also contains constituents such as salts and biocides, which may restrict the growth of microorganisms (Murali Mohan et al. 2013; Cluff et al. 2014; Akob et al. 2015; Kahrilas et al. 2015; Daly et al. 2016). Since the per PW spill volume can vary from small (0.15 m3; fifth percentile) to larger scales (53 m3; 95th percentile), we hypothesise the PW spill volume size is an essential factor in determining the spill impact on microbial communities at PW-contaminated sites, and may ultimately influence natural biodegradation pathways. Although microbial community shifts are critical to assessing the impacts of PW spills, there is currently limited knowledge of fundamental changes in microbial community structures following PW spills into waterways. In this study, we conducted laboratory-simulated late-stage PW spills into field-collected river water from a region of shale oil and gas development. The goals for this study were to (i) determine the impact of differing concentrations of PW on composition and diversity in aquatic microbiota and identify taxa that might be bioindicators of a PW spill, (ii) investigate cell viability kinetics under the influence of PW, and (iii) estimate the biodegradation potential of natural aquatic communities toward organic constituents of the PW. Our study advances our understanding of aquatic microbial community responses in a wide range of spill sizes, while providing fundamental knowledge to assess the fate of PW contaminants in surface releases of variable sizes. METHODS Sample preparation The PW (PW_1) used to determine the impact of differing concentrations of PW on composition and diversity was collected in November 2016. The PW_1 sample returned to the surface at 53 days after initial flowback commenced. The sampling location was from the water/gas separator of a horizontally fractured well (Well ID: 100/12–30-063–21W5) in the Duvernay Formation located near Fox Creek, Alberta, Canada (Fig. S1). The details of the site information and PW_1 collection and transportation methods were described in a previous study (Zhong et al. 2019). The PW_1 sample was stored in sealed pails for 217 days until the experiments began. The freshwater river sample was collected in June 2018 from the Smoky River, which flows through the Duvernay shale oil and gas region (Fig. S1) and is a significant source of water for the makeup of fracturing fluids. Smoky River freshwater was collected in four 5 L sterile glass bottles without headspace, stored in an opaque box with ice, and transported to the University of Alberta within 24 h. Experiments were conducted within 24 h of the arrival of the freshwater sample. Prior to this study, the threshold for observing changes in community composition due to exposure to PW was unclear. The primary goal of this research was to capture dynamics in microbial community composition in freshwater under the effects of PW. To do so, we mixed aliquots of PW_1 and Smoky River freshwater in sterile flasks to total volumes of 100 mL, such that PW_1 consisted of 0.05%, 0.25%, 0.5%, 2.5%, 5%, 25% and 50% of the total volume in the sample series. Notably, we included mixing ratios such as 25% and 50%, which may be less likely to occur in the case of a spill because we aimed to investigate a broad range of concentrations to enhance our understanding of the effects of PW on microbial communities. We also aimed to explore potentially extreme conditions. Pure Smoky River freshwater and pure PW_1 were used as control groups. Sterile controls were prepared by autoclaving the mixed samples (5% PW_1 and 25% PW_1) twice for 45 min at 121°C and 100 kPa. Experiments were conducted in duplicate and all samples were loosely covered with aluminum foil and shaken at 70 rpm at room temperature for 7 days. Here, we used separate flasks for incubations corresponding to each target sampling day, in order to avoid the perturbation of samples as a function of time. Each sample was filtered through 0.22 µm pore size hydrophilic polypropylene membranes (GE Healthcare Life Sciences, Ontario, Canada) at day 0, day 3 and day 7. The sampling scheme references a previous study that studied aerobic biodegradation of synthetic hydraulic fracturing fluids mixed in the laboratory (Kekacs et al. 2015), which itself conformed to the OECD 301 methods (OECD 1992) for studying biodegradation. The filtered membranes were stored at -20°C until DNA extractions. The filtered fluids were stored at 4°C for dissolved organic carbon (DOC) measurements. Cell viability tests Cell viability tests were conducted to determine how cells react to PW as a function of time. The live and dead cells were counted using the Live/Dead BacLight Viability kit (Life Technologies, Ontario, Canada). Due to the field sampling limitations, we were not able to obtain sufficient sample volumes for both the molecular analysis and cell viability tests. To address this issue, we used a PW (PW_2) for the cell viability tests that was as geochemically similar to the PW sample for molecular analysis (PW_1). PW_2 was collected from the same wellpad as PW_1 in September 2016. Cell counting was conducted within 24 h of the sample arrival. We mixed 10% by volume of PW_2 to two additional sources of freshwater (one sample is from a water storage impoundment near the fractured well and the other is from the North Saskatchewan River) and monitored cell viability for 1 month (0, 1 h, 6 h, day 1, day 3, day 7 and day 25). The incubation conditions of the cell viability tests were the same as those used for the molecular experiments. The mixing ratio and temporal schedule aimed at capturing the changes in cell health under the effect of PW_2, as well as cell recovery time afterwards. We aim to provide information about the status of cells when they are exposed to a medium-high range of PW. The results are not intended to be directly comparable with the 16S rRNA gene-based analyses. The detailed methods for live/dead cell counting were presented in our previous study (Zhong et al. 2019). Briefly, 15 randomly selected fields of ‘live’ cells and ‘dead’ cells were counted at 358 × magnification on a Leica DMRXA epifluorescence microscope equipped with fluorescein isothiocyanate (FITC) and rhodamine fluorescence filters. The live cell proportions were calculated by the live cells relative to the total cells per microscope field from the 15 observed microscope fields. The live cells per mL was calculated from the live cells per microscope field by multiplying the conversion factor 1670 from the 15 observed microscope fields. A Student's t-test was used to analyse the statistical difference of the live cell numbers and live cell proportions between treatment groups and their control groups at each observed time point. ANOVA analysis was used to test if mixing 10% PW_2 had significant effects on the numbers and ratios of live cells compared with their controls. Chemical analyses All samples were stored at 4°C until chemical analyses. The chemistry of the PW and Smoky River freshwater samples was characterised, including pH, total dissolved solids (TDS), DOC, total nitrogen (TN), major cations and anions. The TDS was determined by weighing the residual solids after evaporating 10 mL of fluid at 200°C in triplicate. Briefly, cations were measured using an Agilent 8800 inductively coupled plasma mass spectrometer (Agilent Technologies, CA, USA) (Zhong et al. 2019). Anions were measured using DX-600 ion chromatography (ThermoFisher Scientific, MA, USA) and a SmartChem Discrete Wet Chemistry Analyzer, Model 200 (Westco Scientific, CT, USA) (Tabatabai and Frankenberger 1996; Westco Scientific 2007). The measurements of DOC and TN were achieved using a combustion catalyst method with a TOC-V CHS/CSN Total Organic Carbon Analyzer (Shimadzu Corporation, Kyoto, Japan) (Shimadzu Corporation 2001). Sequencing of 16S rRNA genes DNA was extracted from the cells concentrated on filter membranes using the FastDNA Spin Kit for Soil (MP Biomedicals, Solon, OH, USA). DNA extracts in duplicate were pooled together before PCR. The PCR primers were F515 (5′-GTGCCAGCMGCCGCGGTAA-3′) and R806 (5′-GGACTACHVGGGTWTCTAAT-3′), which cover the V4 region of the 16S rRNA gene for bacteria and archaea. The PCR reaction using KAPA HiFi HotStart Ready Mix (Fisher Scientific, Ontario, Canada) began with a 3 min initial denaturation (95°C) followed by 35 cycles of 30 s denaturation (95°C), 30 s primer annealing (55°C) and 30 s extension (72°C), and a final 5 min extension (72°C). The PCR amplicons were submitted to The Applied Genomics Core Sequencing Facility at the University of Alberta for Illumina MiSeq paired-end sequencing. Of note, PCR amplicons were not successfully obtained from abiotic controls templates and the pure PW_1 templates, likely due to low DNA concentrations. Bioinformatics and statistics Raw data were processed following QIIME2 version 2018.6 standard operating procedure (https://qiime2.org/). Briefly, non-chimeric sequences were passed through quality filtering using the DADA2 pipeline implemented in QIIME2 (Table S1). The filtered sequences were aligned to amplicon sequence variants (ASVs) features and further assigned to different taxonomic levels using the q2-feature-classifier, which was trained on Greengenes (version 13.8) at a 99% similarity threshold. The quality-controlled sequences were then manipulated and visualised using R (version 3.5.1) (Wickham 2009; R Core Team 2018). The datasets were rarefied to an even depth of 51 308 sequences in R to conduct alpha- and beta-diversity analyses of the microbial communities (Table S1). Firstly, we conducted beta-diversity analysis and taxonomic analysis to group microbial community shifts by PW mixing ratios. For beta-diversity analyses, a non-metric multidimensional scaling (NMDS) ordination was performed based on the Bray-Curtis distance. The 95% confident intervals of distinguished clusters were identified using the vegan package in R (Oksanen et al. 2018). The envfit function implemented in the vegan package was used to correlate the 10 most abundant genera (the relative abundance of sequences) of the entire datasets to the sample dissimilarity on the NMDS ordination (Oksanen et al. 2018). PERMANOVA was used to investigate the significance (P < 0.05) of the PW_1 proportion on beta diversity. The alpha diversity of each sample was assessed by using the observed numbers of ASVs, the Chao1 Richness Index, the Inverse Simpson's Index and the Shannon Diversity Index implemented in Phyloseq in R (McMurdie and Holmes 2013). The samples at each sampling day were grouped by mixing ratios, which showed similar microbial community responses based on beta-diversity analysis (similarity in ordination space) and taxonomic analysis. ANOVA analysis was used to test whether the PW_1 proportion effect was significant (P < 0.05) on alpha diversity. Tukey's test was used for post-hoc analysis after the ANOVA analysis, regarding whether the differences of alpha-diversity indices between different PW_1 groupings (grouped by similarity in ordination space) were significant (P < 0.05). Similarly, changes of DOC concentrations were represented by the three mixing ratio groupings, and the results of different groupings were compared using ANOVA analysis combined with Tukey's test. The reads have been submitted to the National Center for Biotechnology Information Sequence Read Archive (BioProject: PRJNA593077). Random forest modelling Random forest modelling consists of a large number of decision trees that operate as an ensemble. This approach prevents overfitting of a classification model by averaging multiple classification models together. Here, the genera produced by sequencing were used as variables in the random forest. Important bioindicators in the NMDS clusters were identified by the randomForest package implemented in R (Breiman et al. 2018). For the random forest modelling, samples at day 0 (start points) were excluded from the total dataset and were used as reference points for changes in subsequent samples. As a default, randomForest used 2/3 of the data to construct modelling trees and the remaining 1/3 of the data to test model error. Each random forest model was set to create 500 decision trees. The model was repeated 1000 times from random sampling to model execution. The 20 most important predictors were determined based on the average values of the mean decrease in the Gini Index after 1000 runs of the random forest models were completed. The Gini Index is a measure of how each variable contributes to homogeneity of the nodes and leaves in the resulting random forests, from 0 (homogeneous) to 1 (heterogeneous). The genera with larger Gini Index values were more likely to be variables that separate the targeted groups. RESULTS AND DISCUSSION Chemical characterisation of PWs and Smoky River freshwater The chemical analyses indicated lower pH and higher concentrations of solutes and nutrients in PW than for Smoky River freshwater (Table 1); therefore, even small PW releases may lead to large changes in freshwater chemistry. The pH of Smoky River freshwater was 7.2, while the pH of PW_1 and PW_2 was 4.1 and 4.7, respectively. Reduced pH was a major driver of microbial community changes in hydraulic fracturing impacted streams (Ulrich et al. 2018). Additionally, the TDS, DOC and TN of PWs were all significantly higher than the Smoky River freshwater. TDS concentrations were 219 037 mg L−1 in PW_1 and 216 637 mg L−1 in PW_2. TDS was dominated primarily by sodium and chloride, which were derived from the shale formations (Table S2). The results suggest that the inorganic components of the two PW samples were similar. Salt may be one of the important limiting factors on cell biomass, diversity and the degradation potential of the microbial communities (Davis, Struchtemeyer and Elshahed 2012; Murali Mohan et al. 2013; Cluff et al. 2014; Kekacs et al. 2015; Mouser et al. 2016). DOC concentrations were 85.3 mg L−1 in PW_1 and 200.8 mg L−1 in PW_2, while DOC concentrations were 13.9 mg L−1 in Smoky River freshwater. TN concentrations between the two PW samples were also similar. TN concentrations were 427.6 mg/L and 471.0 mg/L in PW_1 and PW_2, respectively, which were three orders of magnitude higher than for Smoky River freshwater. DOC and TN could originate from several sources, including fracturing chemicals, reservoir hydrocarbons and the injected surface water. In previous untargeted organic analyses of these PW samples at an earlier flowback time from the same wellpad, we demonstrated that the major organic species were fracturing chemicals that included PEGs with 5–25 ethylene oxide units, the biocide alkyldimethylbenzylammonium chloride (ADBAC), and a series of petrogenic compounds such as fluorene and phenanthrene (He et al. 2018). Biocides are one of the major chemical additives used in hydraulic fracturing, and the presence of biocide in hydraulic fracturing fluids may affect the overall performance of biodegradation efforts when a spill occurs (McLaughlin, Borch and Blotevogel 2016). In our samples, ADBAC was below the detection limit. The decrease in biocide concentrations is likely caused by dilution from the formation water or their chemical decomposition and transformation (Kahrilas et al. 2015, 2016). Ammonium was the dominant species in the TN, and the ammonium concentrations were similar to those found in PW produced by hydraulic fracturing in the Marcellus and Fayetteville Formations (up to 420 mg/L) (Harkness et al. 2015). The sources of ammonium remain to be further studied. Based on the previous studies, ammonium in the PW could be associated with the fracturing chemicals such as the breakers (e.g. ammonium persulfate) (Luek et al. 2018) and clay stabilisers (e.g. tetramethyl ammonium chloride) (Butkovskyi et al. 2017). The breakers allow a delayed break down of the gel polymer chains, and the clay stabilisers are used to prevent the swelling of clay particles in reaction to water-base hydraulic fracturing fluids. Besides this, ammonium is also likely to be leached from ammonium-containing clays, evaporites and the thermal degradation of organic matter in the shale oil and gas reservoirs (Liu et al. 2012). Table 1. Selected geochemical parameters of Smoky River freshwater and flowback and produced water samples (PW_1 and PW_2); the rest of the chemistry is presented in Table S2. Samples . pH . TDS (mg/L) . DOC (mg/L) . TN (mg/L) . Cl (mg/L) . Na (mg/L) . PW_1 4.1 219 037 85.3 427.6 104 373 62 831 PW_2 4.7 216 637 200.8 471.0 139 820 68 176 Smoky River freshwater 7.2 168 13.9 0.1 1.4 1.8 Samples . pH . TDS (mg/L) . DOC (mg/L) . TN (mg/L) . Cl (mg/L) . Na (mg/L) . PW_1 4.1 219 037 85.3 427.6 104 373 62 831 PW_2 4.7 216 637 200.8 471.0 139 820 68 176 Smoky River freshwater 7.2 168 13.9 0.1 1.4 1.8 TDS: total dissolved solids. DOC: dissolved organic carbon. TN: total nitrogen. Open in new tab Table 1. Selected geochemical parameters of Smoky River freshwater and flowback and produced water samples (PW_1 and PW_2); the rest of the chemistry is presented in Table S2. Samples . pH . TDS (mg/L) . DOC (mg/L) . TN (mg/L) . Cl (mg/L) . Na (mg/L) . PW_1 4.1 219 037 85.3 427.6 104 373 62 831 PW_2 4.7 216 637 200.8 471.0 139 820 68 176 Smoky River freshwater 7.2 168 13.9 0.1 1.4 1.8 Samples . pH . TDS (mg/L) . DOC (mg/L) . TN (mg/L) . Cl (mg/L) . Na (mg/L) . PW_1 4.1 219 037 85.3 427.6 104 373 62 831 PW_2 4.7 216 637 200.8 471.0 139 820 68 176 Smoky River freshwater 7.2 168 13.9 0.1 1.4 1.8 TDS: total dissolved solids. DOC: dissolved organic carbon. TN: total nitrogen. Open in new tab Grouping samples based on beta-diversity analysis The PW_1 groupings of different PW mixing ratios were determined using beta-diversity analysis. NMDS ordination revealed two key clusters (2.5–5% and >5%) with increasing PW_1 proportions, which were significantly separated from the cluster with PW_1 mixing ratios between 0–0.5% (Fig. 1). Of note, the high PW_1 proportion group (>5% PW_1) changed the least from the starting points (all tested mixing ratios at day 0) according to the NMDS ordination. Compared with clusters with higher PW_1 mixing ratios, the data points with <2.5% PW_1 were more heterogenous. PERMANOVA analyses showed that the PW_1 proportion significantly influenced microbial community structure over the 7 days of incubation (P< 0.05). We defined three PW_1 groupings, reflecting the degrees of effects of PW: low PW_1 (<2.5%), intermediate PW_1 (2.5–5%) and high PW_1 proportions (>5%). It is important to note that the community composition of the 0% PW_1 sample changed extensively during the incubation, while those at higher concentrations did not (Fig. 1). By comparison, the time effect on microbial community dynamics was overshadowed by the PW concentration effect. We subsequently used these groups for statistical analyses of DOC changes, microbial community diversity and composition shifts, and random forest modelling. Figure 1. Open in new tabDownload slide Non-metric multidimensional scaling (NMDS) plot (stress: 0.096) showing differences in microbial community composition for freshwater river samples exposed to 0–50% flowback and produced water (PW_1) at day 0, day 3 and day 7. The 10 most abundant genera of the entire community were correlated with the dissimilarity of the data points. Time factors (in days) appear as numbers above the data points. Dashed lines represent the 95% confidence intervals of each group. Figure 1. Open in new tabDownload slide Non-metric multidimensional scaling (NMDS) plot (stress: 0.096) showing differences in microbial community composition for freshwater river samples exposed to 0–50% flowback and produced water (PW_1) at day 0, day 3 and day 7. The 10 most abundant genera of the entire community were correlated with the dissimilarity of the data points. Time factors (in days) appear as numbers above the data points. Dashed lines represent the 95% confidence intervals of each group. Changes in DOC concentration Geochemical analyses of the fluids suggest that the concentration of organics in PWs were higher than in freshwater. To examine biodegradation potential, we tracked changes in DOC concentrations over 7 days for each mixing ratio. However, ANOVA analyses showed that the relative reductions of DOC in all PW_1 groupings were not significant over 7 days. The largest reduction in DOC over 7 days was an average of 17.5% for the 2.5–5% group (a detailed description of the DOC changes is presented in the supplementary data and Fig. S2). Our results suggest no noteworthy or relevant biodegradation of the PW-related organics during the 7 days of incubation. In contrast to our results using the field-collected PW, the DOC reduction for synthetic hydraulic fracturing fluids can be up to 90% within 7 days of incubation (Kekacs et al. 2015), possibly indicating that the compounds in the field-collected PW can be more difficult to biodegrade than those in synthetic hydraulic fracturing fluids. The reason for the poor biodegradability of field-collected PW is not yet clear. Many factors such as the presence of recalcitrant organics and more complex mixtures of compounds may cause less reduction of DOC in field-collected PW (Kekacs et al. 2015; McLaughlin, Borch and Blotevogel 2016). Microbial community shifts The trends of changes in microbial community diversity and compositions within each PW_1 proportion category were consistent; namely, the mixtures with higher volumes of PW_1 tend to have higher microbial richness and diversity after 7 days of incubation (Fig. 2). ANOVA analysis showed that alpha-diversity indices were significantly (P < 0.05) different between the three PW_1 groupings over 7 days. Following 7 days of incubation, the number of observed ASVs and Chao1 Index in the high PW proportion group (>5% PW_1) was significantly higher (P< 0.05) than those with a lower PW_1 proportion. The Shannon Diversity Index in the >5% PW_1 group was significantly higher (P < 0.05) than for the group with <2.5% PW_1. The Inverse Simpson's Index values were significantly higher (P< 0.05) in the 2.5–5% PW_1 mixtures than in the other two PW_1 groupings. The full results of Tukey's tests, which compared the difference between two groups at day 0, day 3 and day 7, are presented in Table S3. The increased diversity with higher proportions of PW shows uneven diversity, with dominance by a relatively small number of genera. Upon exposure to the disturbance of high levels of PW, these dominant freshwater operational taxonomic units (OTUs) are unable to survive, unmasking the hidden diversity present in the rarer biosphere in these samples. Further, some of these surviving bacteria may grow in response to the increased organics in the PW. The combination of removal of dominant species and increased growth leads to higher overall diversity. This follows the concept of the intermediate disturbance hypothesis, which states that the highest biodiversity will be found at intermediate levels of disturbance (Bendix, Wiley and Commons 2017). Figure 2. Open in new tabDownload slide Temporal changes in microbial diversity between flowback and produced water (PW_1) proportion ranges from <2.5%, 2.5–5%, >5% and a control group at day 0, day 3 and day 7, which suggest that increasing PW concentration may increase the overall taxonomic richness and diversity. Taxonomic richness was represented by (A) Observed ASVs Index and (B) Chao1 Richness Index, while taxonomic diversity was represented by (C) Inverse Simpson's Index and (D) Shannon Diversity Index. The significant thresholds are P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***). Figure 2. Open in new tabDownload slide Temporal changes in microbial diversity between flowback and produced water (PW_1) proportion ranges from <2.5%, 2.5–5%, >5% and a control group at day 0, day 3 and day 7, which suggest that increasing PW concentration may increase the overall taxonomic richness and diversity. Taxonomic richness was represented by (A) Observed ASVs Index and (B) Chao1 Richness Index, while taxonomic diversity was represented by (C) Inverse Simpson's Index and (D) Shannon Diversity Index. The significant thresholds are P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***). The genus Flavobacterium within the phylum Bacteroidetes is known to be prevalent in freshwater environments (Bernardet and Bowman 2006) and was the most abundant bacterium across all the samples at day 0. They consistently constituted the largest proportion of the microbial community in mixtures containing <2.5% PW_1 throughout 7 days of incubation (Fig. 3). The genera Methylotenera and Caulobacter were also a higher fraction of sequences than in other genera in mixtures containing <2.5% PW_1. Compared with the pure freshwater sample, the trends of microbial community dynamics were similar in samples with low PW_1 mixing ratios over 7 days (Fig. 3). Previously characterised members of these genera have been reported to use glucose and methylamine in natural aquatic environments (Wright and Cain 1969; Entcheva-Dimitrov and Spormann 2004; Bernardet and Bowman 2006; Kalyuzhnaya et al. 2010). Our results suggest that relatively low PW concentrations may not dramatically influence the freshwater community. This is likely because the concentrations of PW_1 were too low for there to be a toxicity effect on the community (e.g. pH >∼7.1, salinity <∼5640 mg/L, DOC <∼15.7 mg/L; values based on calculation of the PW proportion), so the indigenous microorganisms were still present. Figure 3. Open in new tabDownload slide Three-factor plot showing temporal changes of the relative abundance in the 10 most abundant genera (y-axis) as a function of time and flowback and produced water (PW_1) mixing ratio. The PW_1 beta-diversity groups (<2.5%, 2.5–5%, >5%) are labelled on the lower x-axis. Day 0, 3 and 7 exposures are labelled on the higher x-axis and are split into three sub-plots, one for each time point. The relative abundance (%) of a genus is represented by the bubble size, and colours on the y-axis represent the phyla of the 10 genera. Each bubble is represented by a data point (the detailed description of each data point and their corresponding values of relative abundance are presented in Figure S3 and Table S4). Figure 3. Open in new tabDownload slide Three-factor plot showing temporal changes of the relative abundance in the 10 most abundant genera (y-axis) as a function of time and flowback and produced water (PW_1) mixing ratio. The PW_1 beta-diversity groups (<2.5%, 2.5–5%, >5%) are labelled on the lower x-axis. Day 0, 3 and 7 exposures are labelled on the higher x-axis and are split into three sub-plots, one for each time point. The relative abundance (%) of a genus is represented by the bubble size, and colours on the y-axis represent the phyla of the 10 genera. Each bubble is represented by a data point (the detailed description of each data point and their corresponding values of relative abundance are presented in Figure S3 and Table S4). Compared with the low PW_1 proportion group (<2.5% PW_1), a pronounced influence on taxonomy compositions began at 2.5% PW_1. The relative abundance of Flavobacterium decreased in experiments having PW_1 proportions of 2.5–5% by day 3 (the average of Flavobacterium decreases from 75% to 20% and 9% in 2.5% PW_1 and 5% PW_1 mixtures, respectively) and remained at lower abundance levels at day 7 (29% and 4% in 2.5% PW_1 and 5% PW_1 mixtures, respectively). The genera Pseudomonas, Rheinheimera, Rhizobium and Brevundimonas were significantly (P < 0.05) correlated with the 2.5–5% PW_1 proportion group (Figs 1 and 3). The enrichments of these genera may be related to the organic constituents introduced by PW. Some previously characterised members of these genera are capable of degradation of a wide variety of hydrocarbons and organics found in hydraulic fracturing fluids such as isopropanol and PEGs (Williams and Sayers 1994; Ahmad, Mehmannavaz and Damaj 1997; Chaîneau et al. 1999; Táncsics et al. 2010; Kekacs et al. 2015; Nuria Obradors 2015). The increase in the relative abundance of these genera is consistent with the higher DOC reduction observed for the 2.5–5% PW_1 proportion group. The shift in microbial community composition is likely to benefit biodegradation processes. However, significant reductions in PW organic concentrations may take a greater length of time. Functional analysis through metagenomics in the future may allow us to better understand the role of particular microorganisms in the biodegradation process. Additional details of the relative abundance of the 10 most abundant bacteria are presented in Table S4. The increase of abundant microorganisms (Fig. 3) plus some less abundant microorganisms (comprising less than 1% of the total sequences) such as genera Aquicella, Geobacter, Massilia, Pedobacter, Planctomyces and Sphingomonas was consistent with the increasing diversity in the mixtures with >0.5% PW_1. This shift suggests that the medium to higher concentrations of PW may inhibit the indigenous freshwater species at natural conditions, while providing substrates for more types of microorganisms to grow. Alternatively, higher concentrations of PW may remove abundant members of the community, allowing for detection of rare community members. Halanaerobium was observed in all mixtures following the addition of PW_1, although at low abundance (less than 1% of the total sequences). They are the most prevalent bacteria in the PW and are capable of using PW-related organics (Daly et al. 2016; Zhong et al. 2019). For mixtures at the highest concentrations of PW_1 (25% and 50%) (e.g. pH <∼6.4, salinity >∼54 900 mg/L, DOC >∼32 mg/L; values based on calculation of the PW proportion), the microbial response is heavily restricted. No significant shift was observed in microbial community compositions for both 25% and 50% mixtures after 7 days of incubation (Fig. 3). It has been demonstrated that elevated salinity (>40 000 mg/L) can inhibit the aerobic degradation of hydraulic fracturing fluid chemicals by microbial communities derived from surface aquatic environments over the course of 7 days (Kekacs et al. 2015). The input salinity in the 25% and 50% mixtures is above this threshold. Thus, the community likely did not shift significantly because the growth of all of its members was inhibited by high salinity. The overall abundance of Archaea is less than 1%, but their presence also appears to be associated with the presence of PW_1 chemical constituents in mixtures. Within theA rchaea, ASVs related to the genus Nitrosopumilus increased from 0.03% to 0.23% in both the 25% and 50% mixtures after 7 days of incubation. The previous study showed that certain strains of the genus Nitrosopumilus, such as Nitrosopumilus maritimus, are able to oxidise ammonia (Walker et al. 2010). Thus, this increase may be associated with the high ammonium concentration found in PW_1. The rest of the Archaea genera changed in their relative abundance by less than 0.1% of all reads. Additional results for the Archaea are presented in Table S5. Cell viability kinetics We mixed 10% PW_2 into hydraulic fracturing source water and North Saskatchewan River water and observed changes in live/dead cell numbers and ratios over 1 month. Here, these results are incorporated into this study to further discuss the potential effect of medium-high mixing ratios of PW on the microbial communities. Compared with the live cell status in the natural conditions, ANOVA tests showed that the treatment of adding 10% PW_2 had significant effects on live cell numbers (P < 0.01) and survived cell proportions (P < 0.05) (Fig. 4). Student's t-tests showed that the trends of live cell numbers and live cell proportions in the treatments remained at statistically lower levels within a day compared with the controls (P < 0.05), then fully recovered in the following days. Specifically, for the treatments of hydraulic fracturing source water and North Saskatchewan River water, the live cell ratios were significantly lower than their control groups immediately after adding 10% PW_2 (Fig. 4A). Live cell ratios in hydraulic fracturing source water increased from 8±13% at the very beginning to the peaks of 52±18% at day 7. Similarly, live cell ratios in North Saskatchewan river water increased from 5±4% to the peaks of 43±11% at day 3. By contrast, live cell ratios in the two freshwater controls either remained consistent or decreased over the observation period (Fig. 4A). Live cell numbers in the treatment groups reduced to ∼104 cells/mL at the very beginning, which were one magnitude and two magnitudes lower than their controls, respectively. Ultimately, live cell numbers increased to above 105 cell/mL at day 7 in the hydraulic fracturing source water treatment group, as well as day 3 in the North Saskatchewan River water treatment group. Live cell numbers in both the freshwater controls either remained consistent or decreased over the observation period (Fig. 4B). Figure 4. Open in new tabDownload slide Temporal changes in (A) live cell ratios (n = 15), and (B) live cell numbers (n = 15) over 30 days in experiments that added 10% flowback and produced water (PW_2) to the hydraulic fracturing source water sample (circle) and the North Saskatchewan River water sample (triangle). The treatment group and the control group are represented by green and red symbols, respectively. Figure 4. Open in new tabDownload slide Temporal changes in (A) live cell ratios (n = 15), and (B) live cell numbers (n = 15) over 30 days in experiments that added 10% flowback and produced water (PW_2) to the hydraulic fracturing source water sample (circle) and the North Saskatchewan River water sample (triangle). The treatment group and the control group are represented by green and red symbols, respectively. The results suggest that PW_2 may immediately kill many of the original freshwater live cells, resulting in dead or metabolically inactive cells shortly after exposure. At relatively low concentration, PW_2 could also provide nutrients for certain microorganisms to be enriched in the mixed conditions. As shown previously for the Duvernay Formation, increasing salinity is highly correlated with decreased cell viability in the fluids produced during the first few days of well flowback (Zhong et al. 2019). The viability tests imply that relatively rapid loss of cell viability due to PW effects could lead to less observed changes in microbial community composition and DOC reduction in the high PW_1 proportion group. Of note, we did not aim to correlate changes in cell viability with the changes in community compositions between the two independent experiments (PW_1 and PW_2). How chemical differences in PW_1 and PW_2 influence dynamic changes in microbial communities, such as the viability of cells in differing DOC concentrations, needs further investigation. Microbial community indicators Analytical methods to fully identify chemical constituents of PW are limited, which may impede evaluation of the impacts of PW releases to freshwater (He et al. 2017a). Sequencing technologies coupled with random forest modelling may allow for microorganisms to be additional indicators for the assessment of PW-contaminated water and/or soil (Ulrich et al. 2018). In our random forests models, 429 bacterial genera and 17 archaeal genera were used to predict the three PW_1 groupings (<2.5% PW_1, 2.5–5.0% PW_1 and >5% PW_1) observed by beta-diversity analyses (Fig. 1). The top 20 important predictors based on the Gini Index score, which were generated by random forest modelling, are shown in Fig. 5. The results show that Pseudomonas, Rhizobium, Sediminibacterium and Brevundimonas are important predictors generated by random forest modelling. Genera such as Brevundimonas, Rhizobium and Pseudomonas were significantly correlated with the 2.5–5% PW_1 group (Fig. 1, Table S6), indicating they could be effective indicators in identifying the effects of spills at the intermediate PW_1 proportion group (2.5–5%). Flavobacterium was the most shifted bacterial genus from the low PW_1 proportion group (<2.5% PW_1) to the intermediate 2.5–5% PW_1 group (Fig. 3). The high Gini Index value suggests that a drop in abundance of Flavobacterium may be an important negative indicator of a spill. Compared with the traditional technique that uses a single or a few microbes to be indicators for pollution, random forest modelling allows for numerous genera to be predictors as an ensemble, including those in relatively less abundance. Here, we found that minor genera such as Cellvibrio (1.4% and 2.6% of the total sequences in the 2.5% PW_1 at day 3 and day 7, respectively) and Shewanella (1.8% and 3% of the total sequences in the 5% PW_1 at day 3 and day 7, respectively) could act as effective predictors of PW exposure. Consistently, indicators (e.g. Rhizobium, Pedobacter and Cystobacter) selected in our random forests model were similar at the family level (e.g. Rhizobiaceae, Sphingobacteriaceae and Cystobacteraceae) to indicator organisms found in streams impacted by hydraulic fracturing in Pennsylvania (Ulrich et al. 2018). Additionally, Rhizobium, Cystobacter, Pedobacter, Herminiimonas, Sediminibacterium and Desulfosporosinus were consistently within the top 20 indicators in our model, and were also similar at the order level (e.g. Clostridiales, Rhizobiales, Myxococcales and Sphingobacteriales), organisms which were enriched in an impacted stream near a shale gas disposal facility in central West Virginia (Akob et al. 2016). As machine learning advances, training additional data in future studies show promise in improving the spill classification accuracy and precision. Figure 5. Open in new tabDownload slide The Gini Index generated by random forests showing the 20 most important genera in predicting PW spills. The random forests technique examines a large ensemble of decision trees, which has considered all the genera represented in the 16S rRNA gene-based sequences. The Gini Index is a measure of how each variable contributes to homogeneity of the nodes and leaves in the resulting random forests, from 0 (homogeneous) to 1 (heterogeneous). A genus having a larger Gini Index is more likely to be a variable that separates the targeted groups. Figure 5. Open in new tabDownload slide The Gini Index generated by random forests showing the 20 most important genera in predicting PW spills. The random forests technique examines a large ensemble of decision trees, which has considered all the genera represented in the 16S rRNA gene-based sequences. The Gini Index is a measure of how each variable contributes to homogeneity of the nodes and leaves in the resulting random forests, from 0 (homogeneous) to 1 (heterogeneous). A genus having a larger Gini Index is more likely to be a variable that separates the targeted groups. Environmental implications and future steps Our study demonstrates that microbial community compositions in aquatic environments are sensitive to the scale of a PW spill. Our results suggest that large volume spills (leading to 25+% PW_1 concentrations), while rare, could have considerable impact. The cell viability test, as a separate experiment, implies that the community may lose its ability to adapt and may not be viable after exposure. We found that the degradation rate for organic constituents of field-collected PW is considerably less than those measured in laboratory-synthesised brines, which calls for more studies using real PW and for long-term monitoring of recalcitrant organic pollutants at contaminated sites. Microorganisms such as Pseudomonas, Rheinheimera, Rhizobium and Brevundimonas may serve as key players in remediation processes, which may also be used as biosensors to assess water bodies that have experienced a PW spill. Moreover, through building a low-error random forests model (Table S7), the sets of genera uncovered in 16S rRNA gene analyses show promise as bioindicators to represent changes in aquatic ecosystems due to PW releases into freshwater. In the future, these bioindicators may complement traditional analytical methods such as chemical analyses in assessing the magnitude or severity of a release. Additional work is necessary to identify the site-specific biomarkers, since the components of PW such as salinity and organic compound identities vary by extraction site and by time of flowback. Moreover, more research is needed to determine whether these bioindicators are universal and how long-lasting they are. ACKNOWLEDGEMENTS The Biological Sciences Molecular Biology Service Unit at University of Alberta provided assistance in interpretation of the microbial community analyses. FUNDING This research was supported by an NSERC Collaborative Research and Development grant (CRDPJ 469308–14) to DSA and GGG, with the support of the Encana Corporation. Conflict of interest None declared. REFERENCES Ahmad D , Mehmannavaz R, Damaj M. Isolation and characterization of symbiotic n2-fixing rhizobium meliloti from soils contaminated with aromatic and chloroaromatic hydrocarbons: PAHs and PCBs . Int Biodeterior Biodegrad . 1997 ; 39 : 33 – 43 . Google Scholar Crossref Search ADS WorldCat Akob DM , Cozzarelli IM, Dunlap DS et al. . Organic and inorganic composition and microbiology of produced waters from pennsylvania shale gas wells . Appl Geochemistry . 2015 ; 60 : 116 – 25 . Google Scholar Crossref Search ADS WorldCat Akob DM , Mumford AC, Orem W et al. . Wastewater disposal from unconventional oil and gas development degrades stream quality at a west virginia injection facility . Environ Sci Technol . 2016 ; 50 : 5517 – 25 . Google Scholar Crossref Search ADS PubMed WorldCat Alessi DS , Zolfaghari A, Kletke S et al. . Comparative analysis of hydraulic fracturing wastewater practices in unconventional shale development: water sourcing, treatment and disposal practices . Can Water Resour J . 2017 ; 42 : 105 – 21 . Google Scholar Crossref Search ADS WorldCat Barbot E , Vidic NS, Gregory KB et al. . Spatial and temporal correlation of water quality parameters of produced waters from Devonian-age shale following hydraulic fracturing . Environ Sci Technol . 2013 ; 47 : 2562 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat Bendix J , Wiley JJ, Commons MG. Intermediate disturbance and patterns of species richness . Phys Geogr . 2017 ; 38 : 393 – 403 . Google Scholar Crossref Search ADS WorldCat Bernardet J , Bowman JP. The genus flavobacterium . Prokaryotes . 2006 ; 7 : 481 – 531 . Google Scholar Crossref Search ADS WorldCat Blewett TA , Delompré PLM, He Y et al. . Sublethal and reproductive effects of acute and chronic exposure to flowback and produced water from hydraulic fracturing on the water flea daphnia magna . Environ Sci Technol . 2017a ; 51 : 3032 – 9 . Google Scholar Crossref Search ADS WorldCat Blewett TA , Weinrauch AM, Delompré PLM et al. . The effect of hydraulic flowback and produced water on gill morphology, oxidative stress and antioxidant response in rainbow trout (Oncorhynchus mykiss) . Sci Rep . 2017b ; 7 : 46582 . Google Scholar Crossref Search ADS WorldCat Brantley SL , Yoxtheimer D, Arjmand S et al. . Water resource impacts during unconventional shale gas development: the Pennsylvania experience . Int J Coal Geol . 2014 ; 126 : 140 – 56 . Google Scholar Crossref Search ADS WorldCat Breiman L , Cutler A, Liaw A et al. . Breiman and Cutler's Random Forests for Classification and Regression . 2018 . https://www.rdocumentation.org/packages/randomForest/versions/4.6-14. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Butkovskyi A , Bruning H, Kools SAE et al. . Organic pollutants in shale gas flowback and produced waters: identification, potential ecological impact, and implications for treatment strategies . Environ Sci Technol . 2017 ; 51 : 4740 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat Chaîneau CH , Morel J, Dupont J et al. . Comparison of the fuel oil biodegradation potential of hydrocarbon-assimilating microorganisms isolated from a temperate agricultural soil . Sci Total Environ . 1999 ; 227 : 237 – 47 . Google Scholar Crossref Search ADS PubMed WorldCat Cluff MA , Hartsock A, Macrae JD et al. . Temporal changes in microbial ecology and geochemistry in produced water from hydraulically fractured marcellus shale gas wells . Environ Sci Technol . 2014 ; 48 : 6508 – 17 . Google Scholar Crossref Search ADS PubMed WorldCat Cozzarelli IM , Skalak KJ, Kent DB et al. . Signatures and effects of an oil and gas wastewater spill in The Williston Basin, North Dakota . Sci Total Environ . 2017 ; 579 : 1781 – 93 . Google Scholar Crossref Search ADS PubMed WorldCat Daly RA , Borton MA, Wilkins MJ et al. . Microbial metabolisms in a 2.5-km-deep ecosystem created by hydraulic fracturing in shales . Nat Microbiol . 2016 ; 1 : 16146 . Google Scholar Crossref Search ADS PubMed WorldCat Davis JP , Struchtemeyer CG, Elshahed MS. Bacterial communities associated with production facilities of two newly drilled thermogenic natural gas wells in the barnett shale (Texas, USA) . Microb Ecol . 2012 ; 64 : 942 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat Engle MA , Cozzarelli IM, Smith BD. USGS Investigations of Water Produced During Hydrocarbon Reservoir Development . 2014 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Entcheva-Dimitrov P , Spormann AM. Dynamics and control of biofilms of the oligotrophic bacterium caulobacter crescentus . J Bacteriol . 2004 ; 186 : 8254 – 66 . Google Scholar Crossref Search ADS PubMed WorldCat Ferrer I , Thurman EM. Analysis of hydraulic fracturing additives by LC/Q-TOF-MS . Anal Bioanal Chem . 2015 ; 407 : 6417 – 28 . Google Scholar Crossref Search ADS PubMed WorldCat Flynn SL , Gunten K Von, Warchola T et al. . Characterization and implications of solids associated with hydraulic fracturing flowback and produced water from the duvernay formation . Environ Sci Process Impacts . 2019 ; 21 : 242 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat Folkerts EJ , Blewett TA, He Y et al. . Alterations to juvenile zebrafish (Danio rerio) swim performance after acute embryonic exposure to sub-lethal exposures of hydraulic fracturing flowback and produced water . Aquat Toxicol . 2017b ; 193 : 50 – 9 . Google Scholar Crossref Search ADS WorldCat Folkerts EJ , Blewett TA, He Y et al. . Cardio-respirometry disruption in zebrafish (Danio rerio) embryos exposed to hydraulic fracturing flowback and produced water . Environ Pollut . 2017a ; 231 : 1477 – 87 . Google Scholar Crossref Search ADS WorldCat Gagnon GA , Krkosek W, Anderson L et al. . Impacts of hydraulic fracturing on water quality: a review of literature, regulatory frameworks and an analysis of information gaps . Environ Rev . 2016 ; 24 : 122 – 31 . Google Scholar Crossref Search ADS WorldCat Goss G , Alessi D, Allen D et al. . Unconventional Wastewater Management: A Comparative Review and Analysis of Hydraulic Fracturing Wastewater Management Practices across Four North American Basins . 2015 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Harkness JS , Dwyer GS, Warner NR et al. . Iodide, bromide, and ammonium in hydraulic fracturing and oil and gas wastewaters: environmental implications . Environ Sci Technol . 2015 ; 49 : 1955 – 63 . Google Scholar Crossref Search ADS PubMed WorldCat He Y , Flynn SL, Folkerts EJ et al. . Chemical and toxicological characterizations of hydraulic fracturing flowback and produced water . Water Res . 2017a ; 114 : 78 – 87 . Google Scholar Crossref Search ADS WorldCat He Y , Folkerts EJ, Zhang Y et al. . Effects on biotransformation, oxidative stress, and endocrine disruption in rainbow trout (Oncorhynchus mykiss) exposed to hydraulic fracturing flowback and produced water . Environ Sci Technol . 2017b ; 51 : 940 – 7 . Google Scholar Crossref Search ADS WorldCat He Y , Sun C, Zhang Y et al. . Developmental toxicity of the organic fraction from hydraulic fracturing flowback and produced waters to early life stages of zebrafish (Danio rerio) . Environ Sci Technol . 2018 ; 52 : 3820 – 30 . Google Scholar Crossref Search ADS PubMed WorldCat Hossack BR , Smalling KL, Anderson CW et al. . Effects of persistent energy-related brine contamination on amphibian abundance in national wildlife refuge wetlands . Biol Conserv . 2018 ; 228 : 36 – 43 . Google Scholar Crossref Search ADS WorldCat Kahrilas GA , Blotevogel J, Corrin ER et al. . Downhole transformation of the hydraulic fracturing fluid biocide glutaraldehyde: implications for flowback and produced water quality . Environ Sci Technol . 2016 ; 50 : 11414 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat Kahrilas GA , Blotevogel J, Stewart PS et al. . Biocides in hydraulic fracturing fluids: a critical review of their usage, mobility, degradation, and toxicity . Environ Sci Technol . 2015 ; 49 : 16 – 32 . Google Scholar Crossref Search ADS PubMed WorldCat Kalyuzhnaya MG , Beck DAC, Suciu D et al. . Functioning in situ: gene expression in methylotenera mobilis in its native environment as assessed through transcriptomics . ISME J . 2010 ; 4 : 388 – 98 . Google Scholar Crossref Search ADS PubMed WorldCat Kekacs D , Drollette BD, Brooker M et al. . Aerobic biodegradation of organic compounds in hydraulic fracturing fluids . Biodegradation . 2015 ; 26 : 271 – 87 . Google Scholar Crossref Search ADS PubMed WorldCat Kondash A , Vengosh A. Water footprint of hydraulic fracturing . Environ Sci Technol Lett . 2015 ; 2 : 276 – 80 . Google Scholar Crossref Search ADS WorldCat Liu Q , Zhijun J, Jianfa C et al. . Origin of nitrogen molecules in natural gas and implications for the high risk of N2 exploration in tarim basin, NW China . J Pet Sci Eng . 2012 ; 81 : 112 – 21 . Google Scholar Crossref Search ADS WorldCat Luek JL , Harir M, Schmitt-Kopplin P et al. . Temporal dynamics of halogenated organic compounds in Marcellus Shale flowback . Water Res . 2018 ; 136 : 200 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat Maloney KO , Baruch-mordo S, Patterson LA et al. . Unconventional oil and gas spills: materials, volumes, and risks to surface waters in four states of the U.S . Sci Total Environ . 2017 ; 581-582 : 369 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat McLaughlin MC , Borch T, Blotevogel J. Spills of hydraulic fracturing chemicals on agricultural topsoil: biodegradation, sorption, and co-contaminant interactions . Environ Sci Technol . 2016 ; 50 : 6071 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat McMurdie PJ , Holmes S. Phyloseq : an R package for reproducible interactive analysis and graphics of microbiome census data . PLoS One . 2013 ; 8 : e61217 . Google Scholar Crossref Search ADS PubMed WorldCat Mouser PJ , Borton M, Darrah TH et al. . Hydraulic fracturing offers view of microbial life in the deep terrestrial subsurface . FEMS Microbiol Ecol . 2016 ; 92 : fiw166 . Google Scholar Crossref Search ADS PubMed WorldCat Murali Mohan A , Hartsock A, Bibby KJ et al. . Microbial community changes in hydraulic fracturing fluids and produced water from shale gas extraction . Environ Sci Technol . 2013 ; 47 : 13141 – 50 . Google Scholar Crossref Search ADS PubMed WorldCat Nuria Obradors JA . Efficient biodegradation of high-molecular-weight polyethylene glycols by pure cultures of Ps eudomonas stutzeri . Appl Environ Microbiol . 2015 ; 57 : 2 – 8 . OpenURL Placeholder Text WorldCat OECD C . Organisation for Economic Co-Operation and Development (OECD) 301 Methods . 1992 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Oksanen AJ , Blanchet FG, Friendly M et al. . Community Ecology Package ‘Vegan’ . 2018 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Orem W , Varonka M, Crosby L et al. . Applied geochemistry organic geochemistry and toxicology of a stream impacted by unconventional oil and gas wastewater disposal operations . Appl Geochemistry . 2017 ; 80 : 155 – 67 . Google Scholar Crossref Search ADS WorldCat Parker KM , Zeng T, Harkness J et al. . Enhanced formation of disinfection byproducts in shale gas wastewater-impacted drinking water supplies . Environ Sci Technol . 2014 ; 48 : 11161 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat Patterson LA , Konschnik KE, Wiseman H et al. . Unconventional oil and gas spills: risks, mitigation priorities, and state reporting requirements . Environ Sci Technol . 2017 ; 51 : 2563 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat Preston TM , Anderson CW, Thamke JN et al. . Predicting attenuation of salinized surface- and groundwater-resources from legacy energy development in the prairie pothole region . Sci Total Environ . 2019 ; 690 : 522 – 33 . Google Scholar Crossref Search ADS PubMed WorldCat Prosser JI , Bohannan BJM, Curtis TP et al. . The role of ecological theory in microbial ecology . Nat Rev Microbiol . 2007 ; 5 : 384 – 92 . Google Scholar Crossref Search ADS PubMed WorldCat R Core Team . R: A Language and Environment for Statistical Computing . 2018 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Shimadzu Corporation . Total Organic Carbon Analyzer TOC-V User Manual . 2001 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Siddique T , Penner T, Klassen J et al. . Microbial communities involved in methane production from hydrocarbons in oil sands tailings . Environ Sci Technol . 2012 ; 46 : 9802 – 10 . Google Scholar Crossref Search ADS PubMed WorldCat Smalling KL , Anderson CW, Honeycutt RK et al. . Associations between environmental pollutants and larval amphibians in wetlands contaminated by energy-related brines are potentially mediated by feeding traits . Environ Pollut . 2019 ; 248 : 260 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat Tabatabai MA , Frankenberger WT. Ion Chromatography: Methods of Soil Analysis, Part 3 - Chemical Methods . Sparks DL (ed). 1996 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Táncsics A , Szabó I, Baka E et al. . Investigation of catechol 2,3-dioxygenase and 16s rrna gene diversity in hypoxic, petroleum hydrocarbon contaminated groundwater . Syst Appl Microbiol . 2010 ; 33 : 398 – 406 . Google Scholar Crossref Search ADS PubMed WorldCat U.S. Environmental Protection Agency . Review of State and Industry Spill Data: Characterization of Hydaulic Fracturing-Related Spills . 2015 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ulrich AC , Guigard SE, Foght JM et al. . Effect of salt on aerobic biodegradation of petroleum hydrocarbons in contaminated groundwater . Biodegradation . 2009 ; 20 : 27 – 38 . Google Scholar Crossref Search ADS PubMed WorldCat Ulrich N , Kirchner V, Drucker R et al. . Response of aquatic bacterial communities to hydraulic fracturing in northwestern Pennsylvania: a five-year study . Sci Rep . 2018 ; 8 : 1 – 12 . Google Scholar Crossref Search ADS PubMed WorldCat Vidic RD , Brantley SL, Vandenbossche JM et al. . Impact of shale gas development on regional water quality . Science (80-) . 2013 ; 340 : 1235009 . Google Scholar Crossref Search ADS WorldCat Walker CB , De Torre JR, Klotz MG et al. . Nitrosopumilus maritimus genome reveals unique mechanisms for nitrification and autotrophy in globally distributed marine crenarchaea . Proc Natl Acad Sci USA . 2010 ; 107 : 8818 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat Wang N , Kunz JL, Cleveland D et al. . Biological effects of elevated major ions in surface water contaminated by a produced water from oil production . Arch Environ Contam Toxicol . 2019 ; 76 : 670 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat Westco Scientific . SmartChem 200 Method 410-200B . 2007 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wickham H . Ggplot2 Elegant Graphics for Data Analysis . 2009 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Williams PA , Sayers JR. The evolution of pathways for aromatic hydrocarbon oxidation in pseudomonas . Biodegradation . 1994 ; 5 : 195 – 217 . Google Scholar Crossref Search ADS PubMed WorldCat Wright KA , Cain RB. Microbial formation of methylamine from 4-carboxy-1-methylpyridinium chloride, a photolytic product of paraquat . Soil Biol Biochem . 1969 ; 1 : 5 – 14 . Google Scholar Crossref Search ADS WorldCat Zhong C , Li J, Flynn SL et al. . Temporal changes in microbial community composition and geochemistry in flowback and produced water from the duvernay formation . ACS Earth Sp Chem . 2019 ; 3 : 1047 – 57 . Google Scholar Crossref Search ADS WorldCat © FEMS 2020. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Response of aquatic microbial communities and bioindicator modelling of hydraulic fracturing flowback and produced water
JF - FEMS Microbiology Ecology
DO - 10.1093/femsec/fiaa068
DA - 2020-05-01
UR - https://www.deepdyve.com/lp/oxford-university-press/response-of-aquatic-microbial-communities-and-bioindicator-modelling-POU0fiwFf9
VL - 96
IS - 5
DP - DeepDyve
ER -