Interstitial water microbial communities as an indicator of microbial denitrifying capacity in wood-chip bioreactors

Interstitial water microbial communities as an indicator of microbial denitrifying capacity in wood-chip bioreactors

Science of the Total Environment 655 (2019) 720–729 Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www...

2MB Sizes 0 Downloads 30 Views

Science of the Total Environment 655 (2019) 720–729

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Interstitial water microbial communities as an indicator of microbial denitrifying capacity in wood-chip bioreactors Soheil Fatehi-Pouladi a,⁎, Bruce C. Anderson a, Brent Wootton b, Mark Button c, Sonja Bissegger d, Lloyd Rozema e, Kela P. Weber d a

Department of Civil Engineering, Queen's University, 58 University Ave., Kingston, ON K7L 3N6, Canada Centre for Advancement of Water and Wastewater Technologies, Fleming College, 200 Albert Street South, Lindsay, ON K9V 5E6, Canada Fipke Laboratory for Trace Element Research, Earth, Environmental and Geographic Sciences, University of British Columbia Okanagan, University Way, Kelowna, BC V1V 1V7, Canada d Department of Chemistry and Chemical Engineering, Royal Military College of Canada, Station Forces, Kingston, ON K7K 7B4, Canada e Aqua Treatment Technologies, 4250 Fly Road, Campden, ON L0R 1G0, Canada b c

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Shortest acclimation period in bioreactor planted with Typha angustifolia. • Interstitial water samples used for microbial assessment. • Direct link between denitrification performance and increased microbial activity • Positive correlation between nitrate reduction and increased denitrifying genes • Longer acclimation period in unplanted reactor than the T. angustifolia reactor

a r t i c l e

i n f o

Article history: Received 30 April 2018 Received in revised form 17 November 2018 Accepted 18 November 2018 Available online 19 November 2018 Keywords: Denitrification Phytotechnology Greenhouse Microbial community nirS Community-level physiological profiling

⁎ Corresponding author. E-mail address: [email protected] (S. Fatehi-Pouladi).

https://doi.org/10.1016/j.scitotenv.2018.11.278 0048-9697/© 2018 Elsevier B.V. All rights reserved.

a b s t r a c t The discharge from food production greenhouses (greenhouse effluent) contains high nutrient and salt concentrations, which, if directly released, can have adverse effects on the environment. Wood-chip bioreactors are increasingly popular passive water treatment systems favoured for their economical denitrification in treating agricultural field tile drainage. Microbial communities are central to denitrification; however little is known about the maturation of microbial communities in wood-chip bioreactors treating greenhouse effluents. In this study, multiple subsurface flow wood-chip bioreactors, each vegetated with a different plant species, together with an unplanted unit, received synthetic greenhouse effluent with elevated nitrate concentrations. The hybrid bioreactors were operated for over 2 years, during which time water samples were collected from the inlet, outlet and within the reactors. The increasing denitrification rate in the bioreactor planted with Typha angustifolia (narrowleaf cattail) correlated with increasing microbial activity and metabolic richness, measured by the carbon utilization patterns in Biolog® EcoPlates. Increased denitrifying gene (nirS) copies (determined by quantitative polymerase chain reaction, qPCR), and near-complete nitrate removal were observed in the T. angustifolia and unplanted reactors after 16 and 23 months of operation respectively. The findings suggested that an acclimation period of at least one year can be expected in unseeded bioreactors planted with T. angustifolia, while bioreactors without vegetation may require a longer time to maximize their denitrification capacity. These results are important for the design and operation of wood-chip bioreactors, which are expected to be more commonly applied in the future. © 2018 Elsevier B.V. All rights reserved.

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

1. Introduction Denitrifying bioreactors are passive treatment systems supplied with an external carbon source, such as wood chips, that are used for nitrate removal from wastewaters. Early studies showed that the available carbon source in the riparian zones near surface waters had an important effect on the denitrification of groundwater that occurred in the soil (e.g. Schipper et al., 1991). These observations led to the innovation of in-situ systems with externally supplied solid carbon for enhanced nitrate removal (e.g. Robertson and Cherry, 1995). In these denitrification systems, since a favourable anoxic environment must be developed, the biologically-available organic carbon (present or externally supplied) serves as an electron donor leading to the reduction of nitrate. This is the key role of the organic carbon in facilitating denitrification. However, the presence of oxygen and aerobic organisms at the start-up period of many systems is an inhibitory condition to successful denitrification. As such, prior to the point when the organic carbon can be utilized for heterotrophic denitrification, these carbon substrates provide another early contribution by serving as an energy source for the aerobic microorganisms and creating the anoxic environment necessary for denitrification (Schipper et al., 2010). It is only after the oxygen has been depleted by the metabolic activity of the aerobic organisms using the organic carbon that denitrification can be started. As reviewed by Addy et al. (2016), there have been many recent studies focused on using denitrifying bioreactors to reduce the adverse impact of the nitrates released from agricultural tile drainage. However, the nitrate concentrations found in the discharges from food production greenhouses (greenhouse effluent) are notably greater than those in a typical agricultural runoff. These high nutrient levels in greenhouse effluent are typically due to the unused portion of the dissolved nutrients that are added to the irrigation water. High nitrate concentrations can lead to hypoxia in surface waters (McIsaac et al., 2001) and cause potential health implications if consumed in drinking water (Knobeloch et al., 2000). Most of the studies on denitrifying bioreactors are comprised of unplanted systems, where employing cover vegetation has received little attention. This is potentially due to the common understanding that plants in passive treatment systems, such as constructed wetlands (CWs) can facilitate oxygen transfer (Kadlec and Wallace, 2009), which in turn might compromise the anoxic conditions necessary for denitrification. However, the oxygen transfer via plant macrophytes in the saturated media of CWs has been shown to be limited (Tanner et al., 2012), and up to 99% nitrate removal from greenhouse effluent (over 200 mg N L−1 incoming nitrate concentration) was recently achieved in a series of pilot-scale vegetated bioreactors (FatehiPouladi et al., 2017). Despite the high performance of denitrifying bioreactors in reducing nitrate loads, the evidence to confirm the presence or function of microbial communities, presumed as the key driver of the reported denitrification, is limited. The few published studies investigating the microbial communities in these bioreactors (e.g. Healy et al., 2015; Warneke et al., 2011) were primarily conducted on unplanted reactors designed for treatment of discharges other than greenhouse effluent. There are a number of microbial assessment methodologies available and employed for investigating treatment wetlands. The quantitative polymerase chain reaction (qPCR) technique is a standard and powerful molecular biological method for assessment of potential function, in which a targeted gene is amplified and measured. For example, the potential number of targeted enzymes, for which the genes are responsible, within different samples can be compared. Other tools developed for assessment of function, such as community level physiological profiling (CLPP), provide an understanding of what the existing microbial community is able to accomplish without directly identifying microbes, or quantifying genes. This is a useful approach, as merely the presence of a certain gene does not necessarily indicate that the

721

environmental conditions are favourable for that gene to be expressed and result in the expected function (Weber, 2016). The microbial assessment of denitrification performance in treatment wetlands and bioreactor systems has typically been conducted by employing qPCR to quantify the copies of 16S rRNA gene and nitrite reductase genes nirK and nirS and the nitrous oxide reductase gene, nosZ (Healy et al., 2015; Warneke et al., 2011; Ligi et al., 2014). The genomic methods, including various types of the qPCR method were the most employed techniques to assess microbial diversity and relative abundance in treatment wetlands in recent years (Faulwetter et al., 2009). In a review conducted by Weber (2016) on microbial assessment methods used in specific research areas between 1988 and 2016, it was reported that the qPCR method was employed in 21 studies focused on nitrogen investigations. In the Community-Level Physiological Profiling (CLPP) method, using Biolog® microplates, the environmental sample is incubated with various types of carbon sources, and the colour change developed through the production of nicotinamide adenine dinucleotide (NADH) over the incubation period provides information regarding metabolic function of the mixed microbial communities (Weber and Legge, 2010). In many recent studies, the parameters derived from the CLPP method have helped researchers to understand the microbial function of treatment wetlands of different scales and designs (e.g. Bissegger et al., 2014; Button et al., 2015; Zhang et al., 2018). In our study, CLPP was employed to determine the metabolic activity of the mixed microbial communities based on their carbon utilization, while qPCR was used to quantify the denitrification genes. Both of these methods can generate data to understand and monitor the potential function of microbial communities and their related activities. To the knowledge of the authors, the use of qPCR for denitrification genes and CLPP for the carbon metabolic profile/activity in tandem has never been done. The organic carbon is an integral element in a denitrification process and often the limiting factor in wastewaters with high nitrate concentrations. As the CLPP method is based on the metabolic activities of the heterotrophic microbial communities using different sources of carbon, our integrated approach of evaluating denitrification systems by quantifying denitrifying genes (via qPCR) and assessing the heterotrophic microbial communities (via CLPP) is a novel and relevant methodology. The aim of this study was to 1: investigate the development of microbial communities in differently vegetated wood-chip bioreactors with no prior seeding by measuring the metabolic function of mixed microbial communities via the CLPP method; 2: characterize the denitrifying capacity of the bacterial community by investigating the denitrifying gene copies present in the bioreactors; and 3: present findings to assist designers and operators of wood-chip bioreactors. Overall, this study investigated the microbial denitrification capacity of vegetated wood-chip bioreactors as a potentially effective mechanism in treating the effluents from greenhouse operations. The research objectives were defined to describe whether the nitrate removal exhibited from the water quality measurements could be explained by the microbial activity assessments. As each of the vegetated bioreactors was planted with a different plant species including halophytes and non-halophytes, the role of the plants in the microbial development of denitrifying bioreactors was also of interest. 2. Materials and methods 2.1. Timeline and reactor design The pilot-scale project was built and operated in the Department of Civil Engineering at Queen's University (Kingston, ON, Canada). Each of the 5 bioreactors was constructed from open-top 220-L barrels (diameter: 56 cm, height: 90 cm) and filled with maple hard wood chips, acquired from Quebec, Canada, in a single 80-cm layer. A vertical PVC tube (inner diameter: 2.54 cm) was positioned in the centre of each

722

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

Fig. 1. a: Schematic of a planted wood-chip reactor with an outlet pipe, used to maintain the reactor's water level; b: vertical sampling platform in the center of the same reactor shown in a; and c: sampling platform supporting 3 interstitial water sampling ports. Reactors were made of black opaque polyethylene barrels (transparent barrel wall is for illustration only).

reactor (planted and unplanted) through the wood-chip media as a platform to support 3 interstitial water sampling tubes (inner diameter: 1.6 mm), which were installed at 20, 40 and 60 cm below the top surface of the wood chips (Fig. 1). The synthetic greenhouse effluent was prepared in the laboratory as per the common greenhouse effluent characteristics reported in the literature and by greenhouse operations in Ontario, Canada (Rozema and Gordon, 2014; Koide and Satta, 2004; Park et al., 2009; Fatehi-Pouladi et al., 2016). Fatehi-Pouladi et al. (2016) summarize typical greenhouse effluent concentrations. The average parameters measured in the synthetic greenhouse influent are shown in Table 1. The total operation was divided into the HL: high loading (average influent NO3–N: 307 mg L−1) and LL: low loading phase (average influent NO3–N: 202 mg L−1). The bioreactors were operated in vertical subsurface-flow (VSSF) mode. The influent was continuously distributed over the surface of

Table 1 Average parameters measured in the greenhouse effluent ± SD (standard deviation). Parameter (unit) NO− 3 –N NO− 2 –N

−1

(mg L ) (mg L−1) NH3–N (mg L−1) TKN–N (mg L−1) −1 PO3– ) 4 –P (mg L −1 SO2− ) 4 –S (mg L Alkalinitya (mg L−1) BOD5 (mg L−1) COD (mg L−1) COD/BOD5 pH a

Alkalinity as CaCO3 to pH 4.5.

High loading

Low loading

307.3 (9.9) 1.8 (0.6) 8.0 (0.8) 9.3 (1.8) 26.2 (3.5) 131.0 (4.1) 8.6 (3.9) 2.8 (1.3) 32.6 (11.0) 12 5.5 (0.5)

201.8 (4.8) 1.1 (0.3) 5.5 (0.4) 6.8 (1.0) 16.8 (2.2) 129.5 (3.9) 5.3 (1.9) 3.1 (1.5) 16.3 (3.8) 5 5.2 (0.6)

each reactor by means of a peristaltic pump (hydraulic loading rate: 12 cm d−1 (122 L d−1 m−2) and a square distribution grid, while the outflow was discharged from the bottom. The 4 vegetated bioreactors together with the unplanted (control) unit were in continuous operation indoors (ambient air temperature: 24 °C) under a 1000 W Metal Halide grow light set to a 16 h/8 h on/off cycle to simulate the sunlight. The project timeline, operational phases and various test dates are presented in Fig. 2. The active plant species changed throughout the operation. The active plant species were grouped as A, B and C. The A selection included softstem bulrush (Schoenoplectus tabernaemontani C.C. Gmel. Palla), switchgrass (Panicum virgatum L.), narrowleaf cattail (Typha angustifolia L.) and Canada wildrye (Elymus canadensis L.). The species present in the B group were S. tabernaemontani and T. angustifolia as P. virgatum and E. canadensis did not survive the experimental conditions. Due to the loss of P. virgatum and E. canadensis in the initial phase, the second phase was continued with only S. tabernaemontani and T. angustifolia (B Group). Prairie cordgrass (Spartina pectinata Bosc ex Link) and saltgrass (Distichlis spicata L. Greene) were then introduced in the final phase. Hence, the C group included S. tabernaemontani, T. angustifolia, S. pectinata and D. spicata when the operation resumed in June 2015.

2.2. Microbial community assessment The interstitial water samples were drawn from the spatial vertical profile using a set of dedicated 50 mL plastic syringes, which were pre-rinsed with deionized water. The collected samples were then transferred to air-tight sterile polypropylene tubes. Samples were consequently transported to the laboratory of the Environmental Sciences Group (ESG) at the Royal Military College of Canada (Kingston, ON, Canada) for further preparation.

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

723

Fig. 2. The operation timeline, loading phases, sampling dates and plant species. CLPP and DNA refer to the events for 4 CLPP and 7 qPCR sample collection and analyses respectively (time intervals not drawn to scale).

2.2.1. CLPP Biolog® EcoPlates (BIOLOG Inc., Hayward CA, USA) with 31 different carbon sources and one blank well in triplicate (96 wells per plate) were inoculated with each water sample at room temperature on the same day the samples were collected. Sample preparation was performed in an aseptic environment using sterile laboratory consumables and 70% ethanol spray for sterilizing the tools and bench tops. 100 μL of the unfiltered samples were pipetted to each well, after which the plates were stored in the dark during the incubation period. The optical densities of the plates were read using a microplate absorbance reader (TECAN Infinite®, Tecan US Inc., Morrisville, NC, USA) after each plate was first shaken by the reader. The optical densities were read at an absorbance of 590 nm at several intervals after inoculation. The first two CLPP tests (CLPP-1 and 2), were performed as per the commonly practiced aerobic protocol where sample collection, preparation, inoculation and incubation procedure were exposed to natural oxygen in air. CLPP-3 and 4 were conducted anaerobically to better preserve the anaerobic microbial communities and inhibit aerobic respiration. Anaerobic CLPP procedure has previously been described and used by Röling et al. (2000). In the present study, nitrogen gas was used to purge oxygen from the sample tubes by de-airing them prior to sample collection. The tubes were then sealed using multi layers of adhesive tape. The plate inoculation was performed in an oxygen-free environment inside a glove box (or a glove bag) purged with a high flow of nitrogen gas. All the laboratory equipment including the samples and plates were stored inside the de-aired chamber while a small continuous flow of nitrogen ensured a low-oxygen environment. The plates, petri dishes and other tools were similarly purged prior to use, and the inoculated plates were sealed using adhesive tape around each plate ensuring the top lids and bottom surfaces remained intact. The plates were stored in the chamber under opaque black covers during the incubation period to prevent light from entering and were only removed temporarily to record their absorbance values. The lids remained taped to the plates during the entire process. The CLPP data analysis, the overall average well colour development (AWCD) and richness were performed according to the previously established methods of Weber et al. (2007) and Weber and Legge (2010). The absorbance readings between 48 and 50 h (for all CLPP samples) were selected as the suitable time point for further analysis. These time points were selected after analyzing the absorbance patterns in the plates over a different range of incubation periods of up to 138 h.

The richness of the microbial community, also described as the metabolic potential of the community, was calculated as a ratio of the number of the carbon sources that were utilized by the inoculated microorganisms, over the total number of carbon sources, presented in percentage (%). The carbon substrates in each Biolog® plate have been grouped in 5 major guilds, namely carbohydrates, carboxylic/acetic acids, polymers, amines/amides and amino acids (Weber and Legge, 2009). Moreover, 8 of the total 31 carbon substrates have recently been identified for their common occurrence in various plant root exudates (Button et al., 2016). 2.2.2. qPCR The expression level of the nitrite reductase genes nirK and nirS was determined by qPCR using the SYBR Green detection system. The primers and programs used for the amplification of nirK and nirS have been previously designed (Ligi et al., 2014). The influent and interstitial water samples were first filtered using nylon membrane filters (0.2 μm pore size, diameter: 47 mm, Millipore, Etobicoke, ON, Canada). Filters with the captured residues were then placed in sterile tubes and stored at −20 °C until the DNA extraction was conducted. The total community DNA extraction was performed by rolling the thawed filters inside the extraction vials provided in the FastDNA SPIN Kit for Soil (MP Biomedicals, Santa Ana, CA, USA), following the manufacturer's protocol. The nucleic acid concentrations were measured by pipetting 1 μL of the extracted samples onto the NanoDrop 2000 spectrophotometer (Fisher Scientific, Toronto, ON, Canada). The samples were analyzed undiluted in duplicates on an Eco™ Real-Time PCR System (Illumina, San Diego, CA, USA) using Maxima SYBR Green Master Mix (Fisher Scientific, Toronto, ON, Canada) and 0.4 μM of forward and reverse primers. The reaction efficiencies calculated by EcoStudy Software (Illumina, San Diego, CA, USA) were 100 ± 10% with an R2 N 0.990. On each plate, a standard curve (0.025 to 2.5 ng) and no template control were run with the samples. The qPCR standard curves were prepared using 6 serial dilutions (25.00, 6.25, 1.56, 0.40, 0.10 and 0.025 ng) of the DNA extracted from thermophilic sludge from Ravensview wastewater treatment plant (Kingston, ON, Canada). PCR product sizes were verified against a 1 kb Plus DNA Ladder (ThermoFisher) using an agarose gel. The gene copy concentrations were derived relative to the standard curves. Each value was normalized for the nucleic acid content of the

724

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

extracted DNA sample. The normalized gene copies of all the samples were then standardized against that of the influent sample to report the gene copy fold-change. It is important to note that gene quantities are not absolute and are relative numbers. 2.2.3. Water quality and statistical analysis With the exception of pH, the water quality characteristics shown in Table 1 and the nitrate concentrations in the influent and outflow grab samples were analyzed at an external laboratory according to Standard Methods (APHA, 2005). The pH levels were measured on-site using an accumet® Model 15 pH meter (Fisher Scientific™, Toronto, ON, Canada). All of the statistical analyses including one-way and two-way ANOVA, as well as the Tukey's and Dunnett's post hoc tests were performed using XLSTAT software (®Addinsoft, New York, NY, USA) and a difference was reported significant when p–value b 0.05. 3. Results and discussion The microbial activity (AWCD) and the richness derived from the CLPPs of the mixed microbial communities were generally highest in the bioreactor planted with S. tabernaemontani in the first CLPP test conducted in the early stages of the experiment (Table 2 and Fig. 3). In the rest of the measurements, the AWCD and richness values were highest in the T. angustifolia reactor. The T. angustifolia reactor was identified as significantly different from the unplanted unit for its metabolic activity within CLPP-1 to 4. The 2-way statistical analysis of the AWCD values with depth (top, middle and bottom) as an independent factor suggested that the reactor depth was not a significant factor in any of the CLPP tests. These depth variations as shown in Fig. 3 made it difficult to identify a consistent pattern in the AWCD values along the vertical profile. The AWCD in the T. angustifolia reactor increased from 0.33 in CLPP1 to 1.08 in CLPP-4 (over 3-times increase), while the denitrification rate of this reactor reached its maximum of 99% (32 g N day−1 m−3) after approximately 16 months of continuous operation (Day 491: between CLPP-3 and 4, Figs. 4.a and 5.a). Although the nitrate removal performance fluctuated over the experimental period; it trended upwards as time progressed (Fig. 4.a). The AWCD values between Days 149 and 530 are not available; however, we would expect these AWCD values to be in line with the overall trends gathered over the extended experimental period. The nitrate limitation caused by this high denitrification performance provided suitable conditions for the anaerobic respiration of the sulfate reducing bacteria, in which sulfate was used as the terminal electron acceptor. The sulfate reducing environment was evidenced by the decreased outflow sulfate concentrations, as well as the perceived sulfide precipitation that led to a high outflow water turbidity (Fatehi-

Pouladi et al., 2017). The increasingly higher AWCD values in the T. angustifolia reactor over time correlated with the higher microbial activities of the developing denitrifying and sulfate reducing bacteria. The denitrification performance of our wood-chip bioreactors during the HL and LL phases (with high and low nitrate loading) was discussed previously (Fatehi-Pouladi et al., 2017). As reported in that study, all the reactors were overloaded by excessive levels of nitrate concentrations during the HL phase, and the bioreactors were only able to reduce nitrate between 30% and 55%. This limited denitrification was attributed to the limited biologically available organics (measured as BOD5 and COD in the greenhouse effluent, displayed in Fig. 6), leached from the wood chips and the plants below-ground biomass. The decrease in the inflow nitrate concentrations (LL phase) then coincided with very high levels of organic carbon in the form of BOD5 that were made available, likely as a result of the vigorous growth of T. angustifolia. This high availability of organics in the water resulted in an average denitrification efficiency of over 88% in the T. angustifolia reactor. The relationship between the nitrate removal and microbial activities and the capacity of denitrifying genes, which are the focus of this paper, were however not discussed in the previous study. The water quality measurements indicated the denitrification rate in the T. angustifolia reactor remained above 30 g N day−1 m−3 (99% NO3– N removal) during a period of at least 9 months after the last CLPP test was performed on Day 531 (Fig. 5.a), suggesting that suitable conditions for successful denitrification continued to be maintained. In the beginning of the experiment, the greatest microbial activity was measured at 0.72 in the S. tabernaemontani bioreactor (Fig. 3.a, CLPP-1). This was concurrent with the highest nitrate reduction of 99% (49 g N day−1 m−3) observed in this reactor throughout the duration of this study. However, the nitrate removal rate in this reactor decreased rapidly as the operation progressed, and the metabolic activity dropped to 0.35 in CLPP-4 as well. The decreased activity and denitrification of this reactor were likely due to the reduction in the supply of organic carbon from wood chips after the initial leaching that deprived the heterotrophic community of the energy required for their metabolism (outflow BOD5 b 2 mg L−1 on Day 559). Of the four CLPP tests, the results of CLPP-4 can be considered the most representative of the matured state of the system, as it followed an anaerobic protocol, and was conducted approximately 17 months after the project start-up. Despite the statistical difference that was reported between the control reactor and all the other units (Table 2), the T. angustifolia reactor was the only unit with considerably higher metabolic activity, while the other reactors showed lower, or in case of S. pectinata, only slightly higher numbers than the control. In addition to the AWCD values, the carbon utilization richness associated with plant root exudates for the T. angustifolia reactor displayed a consistent increase from 37% in CLPP-1 to 81% in CLPP-4 (Fig. 4.b), suggesting that the planted reactor could utilize a higher number of root

Table 2 Mean AWCD values ± SD (standard deviation) and mean total richness (%) in each bioreactor (n: 3 depths) and influent (absorbance readings between 48 h to 50 h). CLPP-1 AWCD

CLPP-2 Rich.

AWCD

(%) Influent Control E. canadensis P. virgatum S. taber1 T. angustifolia S. pectinata D. spicata

0.36 0.55 ± 0.18 a,b 0.55 ± 0.04 a,b 0.33 ± 0.02 b,⁎ 0.72 ± 0.02 a 0.33 ± 0.02 b,⁎ – –

45 48 58 39 64 39 – –

CLPP-3 Rich.

AWCD

(%) 0.36 0.25 ± 0.09 b,c 0.11 ± 0.01 c 0.31 ± 0.08 b 0.36 ± 0.08 a,b 0.54 ± 0.05 a,⁎ – –

Different letters: significant paired difference within each CLPP test (p–value b 0.05). 1 S. tabernaemontani. ⁎ Data with significant difference from the control reactor (p–value b 0.05).

39 35 13 42 42 55 – –

CLPP-4 Rich.

AWCD

(%) 0.20 ± 0.25 0.33 ± 0.01 b,c 0.21 ± 0.08 c 0.30 ± 0.01 b,c 0.43 ± 0.13 a,b 0.55 ± 0.11 a,⁎ – –

29 32 26 45 45 55 – –

Rich. (%)

0.24 ± 0.23 0.51 ± 0.16 b,c – – 0.35 ± 0.06 c 1.08 ± 0.13 a,⁎ 0.64 ± 0.02 b 0.32 ± 0.03 c

42 45 – – 35 74 48 32

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

725

Fig. 3. AWCD values for each species (a to d), control (e) and influent (f) measured by CLPP at different times (absorbance readings between 48 h to 50 h). The planted and control units: AWCD at top, middle and bottom depths. Influent: single sample from the influent tank. Plant representations are for illustration only.

exudate carbon substrates as the bioreactor matured. This indicates that the presence of T. angustifolia and its roots played a role in the maturation of the microbial community that developed within the reactor.

Compared with the other plants employed in this study, T. angustifolia had considerably higher below-ground and aboveground biomass growth rate and developed the largest root biomass

726

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

Fig. 4. Nitrate removal rate, microbial activity and root exudate richness in the wood-chip bioreactors. a: Temporal trend of NO3–N removal rate (left vertical axis) and the same reactor's AWCD (right axis); b: average microbial community richness based on the utilized carbon sources associated with plant root exudates. Error bars represent the standard deviation of 3 depths in both figures.

(determined visually). A positive correlation between the dissolved organic carbon release and plant relative growth rate in CW systems has been suggested before (Zhai et al., 2013). Therefore, the higher microbial activity and root exudate richness (Fig. 4), observed in our T. angustifolia reactor can be attributed to the large rhizosphere and root network, which are typical of T. angustifolia. These results should however be interpreted with caution, as the metabolic activity and root exudate richness values of the influent also ranged from 0.20 to 0.36 and 25% to 62% respectively over the same period. This observation can be an indication of a natural microbial community establishment inside the influent reservoirs. However, compared to the reactors, the lower influent pH (b5.5) and a substantially different environment in the influent reservoir (covered throughout the operation by a plastic sheet) indicate that the communities established in the reservoir were likely very different from the denitrifying and sulfate reducing bacteria in the reactors. The absolute and normalized guild analysis of the reactor and influent samples are provided in supplementary data. The figures included in the supplementary data display how the microbial communities in the influent and reactor samples developed to become different from one another over time.

The nirS gene copies were positively correlated (Pearson's R of 0.331, n = 105, p b 0.002) with the denitrification rate (Fig. 5.b and c). Similar to the mixed microbial activity results, the highest nirS copies were measured in the bioreactor planted with T. angustifolia. The average fold-change of nirS in this reactor revealed a sudden increase in the second half of the operation with an average of 3464, 715 and 679-times increase on Days 517, 601 and 665 respectively (Fig. 7.a). Similar to Healy et al. (2015), who investigated the denitrification genes along their experimental wood-chip columns, we did not observe a consistent trend in the denitrification gene copies with the reactors' depth. The nitrate removal rate and the Ln (nirS fold-change) were positively correlated in the T. angustifolia reactor (Fig. 7.b, R2 = 0.78). A positive correlation between the total copies of nitrite/nitrous oxide reductase genes (nirS, nirK and nosZ) and nitrate removal rate was previously shown in a study of denitrifying bioreactors with carbon-rich media (Warneke et al., 2011), where the reductase gene copies were found to be lower in hard and soft wood chips in comparison with the other carbon substrates such as maize cobs, wheat straws and sawdust. The nirK fold-changes in our study were found to be negligible and hence not reported. However, nirK populations in other denitrifying systems were typically found in abundance, with a nirS/nirK ratio ranging

Fig. 5. Temporal trends for nitrate removal and nirS gene copies for all reactors. a: Nitrate mass removal rates; b: relationship between nirS gene copies and nitrate removal %; c: relationship between nirS gene copies and the nitrate mass removal rate.

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

727

Fig. 6. Influent and effluent BOD5 (a) and COD (b) concentrations.

from 1 to 13 in different environmental conditions (Ligi et al., 2014; Warneke et al., 2011). The low nirK copies in our system can be attributed to the fact that our bioreactors were constructed and operated in a laboratory setting and received synthetic influent in reasonably controlled conditions. Thus, it was expected that the developed microbial communities were not as diverse as reported in other systems. In the T. angustifolia reactor, the higher activities of the mixed microbial communities interpreted from the AWCD values, together with the higher nitrite reductase gene (nirS) copies are strong markers to indicate the elevated denitrification performance in this reactor was caused by microbial pathways involving denitrifying bacteria. However, future work to analyze the absolute quantities of nirS and nirK would be useful to quantify the denitrification rates.

It has been shown that CWs planted with Typha latifolia, another species of the same family, encouraged the growth of nirK containing bacteria, while the addition of the plant's litter could stimulate the growth of the bacteria expressing nirS genes (Chen et al., 2014). As T. angustifolia had the fastest growth rates and the largest root mass, these findings suggest the possibility that the dead and decaying biomass of T. angustifolia in our study might have contributed to the higher overall activity of the bacterial community and increased quantity of the nirS genes. Since the results of the T. angustifolia mesocosm stood out in all the different analyses, the presence of T. angustifolia can be regarded as an important factor in providing the favourable conditions for growth of heterotrophic bacteria.

Fig. 7. Temporal and spatial trends of nitrate removal and nirS gene copies. a: Temporal trend of nirS fold change and nitrate removal efficiency in the T. angustifolia reactor; b: relationship between nitrate removal rate and nirS gene copies.

728

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729

An important observation that ensued was the relatively high nirS copies in the unplanted reactor (Fig. 5.b and c) despite the nitrate maximum removal of 46% around the time when the final DNA sampling was performed (Day 665). It should however be noted that the nitrate reduction in this reactor reached 99% (30 g N day−1 m−3) on Day 699, about one month after the final DNA testing (Fig. 5.a). This delayed high denitrification performance demonstrated the unplanted reactor required a longer acclimation time to attain its full denitrification capacity. Furthermore, the recognizable higher nirS fold-change observed in this reactor was likely captured during the exponential growth phase of its bacterial community, which proved an effective measure predicting the next phase, where denitrification rate was maximized and nitrate concentrations became the limiting growth factor. It was earlier suggested that the substantially higher release of organic carbon (shown by outflow BOD5 and COD measurements) in the T. angustifolia reactor (Fig. 6) was the main driver for the nearcomplete denitrification (Fatehi-Pouladi et al., 2017). The degradation of the wood chips can be a major factor in leaching organic matter that provides the substrate and energy source for the heterotrophic denitrifying bacteria. The appearance of white substances on the surface of the wood chips in our bioreactors could hint at the presence and growth of fungi species as well. Fungi have been shown to contribute to co-denitrification and production of nitrogen gas in soils (Long et al., 2013). In addition to a potentially direct role, the fungi species may have an indirect role by aiding the denitrifying bacteria by degrading the cellulose and lignin in wood chips and making the organic carbon bio-available for the bacteria (Appleford et al., 2008; Porter, 2011). However, further investigations are needed and recommended as future steps to gain deeper insights into the function of the fungi and denitrifying bacteria in enhancing denitrification in direct or indirect ways. Furthermore, although the interstitial water sampling method has previously been used in microbial studies of CWs to avoid destructive core sampling techniques (e.g. Button et al., 2016), the microbial characteristics of the biofilm attached to the wood chips are still of great interest. Assessment of the microbial communities in the attached biofilm is recommended in the future studies in order to provide a basis for comparison with the current investigation of the interstitial water samples. The findings of this study can assist the engineers and greenhouse managers/operators with the start-up time expectations, as related to the design and management of planted wood-based denitrification systems. The long-term nature of the pilot-scale experiments presented in this paper can inform designers about the dynamic nature of wood-chip bioreactors in long-term operations. Further testing is required to determine and confirm the longevity of the release of organics from wood chips and plant roots. Our findings suggest that denitrification can be dynamic, depending on a number of factors, including organic release as a result of root exudates and wood chips. It was also shown that bioreactors vegetated with capable species such as T. angustifolia, can host and develop a more active microbial community than other species such as D. spicata and result in a better denitrification performance with a shorter start-up period. Furthermore, while reactors planted with capable species can be favoured for their faster establishment of functional microbial communities, unplanted systems can still be employed in simpler applications where a longer start-up phase can be accommodated. The important effect of the microbial aspect of these systems on their performance as discussed in this study should be the focus of the future optimization efforts. 4. Conclusions It was demonstrated that high denitrification performance of woodchip bioreactors in treating greenhouse effluent can be linked to the mixed microbial community and bacterial population encoding the denitrifying gene of nirS, thus confirming the biologically mediated denitrification in these bioreactors. The T. angustifolia reactor showed the

shortest acclimation period among all of the reactors, followed by the unplanted unit. These acclimation periods were shown to be related to the microbial development within each system. The presence of T. angustifolia was discussed as the factor that enhanced microbial denitrification, while the unplanted unit demonstrated comparable performance over a longer acclimation period. Acknowledgements Funding: This work was supported by the College-University Idea to Innovation (CUI2I) Grants program of the Natural Sciences and Engineering Research Council of Canada (NSERC) [Award Number: CUI2I/ 437621-2012]. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.scitotenv.2018.11.278. References Addy, K., Gold, A.J., Christianson, L.E., David, M.B., Schipper, L.A., Ratigan, N.A., 2016. Denitrifying bioreactors for nitrate removal: a meta analysis. J. Environ. Qual. 45, 873–881. https://doi.org/10.2134/jeq2015.07.0399. APHA, 2005. Standard Methods for the Examination of Water and Wastewater. 21st ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington, DC, USA. Appleford, J.M., Rodriguez, L.F., Cooke, R.A., Zhang, Y., Kent, A.D., Zilles, J., 2008. Characterization of Microorganisms Contributing to Denitrification in Tile Drain Biofilters in Illinois. https://doi.org/10.13031/2013.24655 (2008 Providence, Rhode Island, June 29–July 2, 2008). Bissegger, S., Rodriguez, M., Brisson, J., Weber, K.P., 2014. Catabolic profiles of microbial communities in relation to plant identity and diversity in free-floating plant treatment wetland mesocosms. Ecol. Eng. 67, 190–197. https://doi.org/10.1016/j. ecoleng.2014.03.063. Button, M., Nivala, J., Weber, K.P., Aubron, T., Müller, R.A., 2015. Microbial community metabolic function in subsurface flow constructed wetlands of different designs. Ecol. Eng. 80, 162–171. https://doi.org/10.1016/j.ecoleng.2014.09.073. Button, M., Rodriguez, M., Brisson, J., Weber, K.P., 2016. Use of two spatially separated plant species alters microbial community function in horizontal subsurface flow constructed wetlands. Ecol. Eng. 92, 18–27. https://doi.org/10.1016/j. ecoleng.2016.03.044. Chen, Y., Wen, Y., Zhou, Q., Vymazal, J., 2014. Effects of plant biomass on denitrifying genes in subsurface-flow constructed wetlands. Bioresour. Technol. 157, 341–345. https://doi.org/10.1016/j.biortech.2014.01.137. Fatehi-Pouladi, S., Anderson, B.C., Wootton, B., Rozema, L., 2016. Evaluation of phytodesalination potential of vegetated bioreactors treating greenhouse effluent. Water 8. https://doi.org/10.3390/w8060233. Fatehi-Pouladi, S., Anderson, B.C., Wootton, B., Rozema, L., 2017. Evaluation of passive reduction of nitrate from greenhouse effluent by planted bioreactors. Water Sci. Technol. Water Supply 17 (6), 1764–1773. Faulwetter, J.L., Gagnon, V., Sundberg, C., Chazarenc, F., Burr, M.D., Brisson, J., Camper, A.K., Stein, O.R., 2009. Microbial processes influencing performance of treatment wetlands: a review. Ecol. Eng. 35, 987–1004. Healy, M., Barrett, M., Lanigan, G., Joo Serrenho, A., Ibrahim, T., Thornton, S., Rolfe, S., Huang, W., Fenton, O., 2015. Optimizing nitrate removal and evaluating pollution swapping trade-offs from laboratory denitrification bioreactors. Ecol. Eng. 74, 290–301. https://doi.org/10.1016/j.ecoleng.2014.10.005. Kadlec, R.H., Wallace, S., 2009. Treatment Wetlands. 2nd ed. CRC Press, Taylor & Francis Group, Boca Raton, FL, USA. Knobeloch, L., Salna, B., Hogan, A., Postle, J., Anderson, H., 2000. Blue babies and nitratecontaminated well water. Environ. Health Perspect. 108 (7), 675–678. Koide, S., Satta, N., 2004. Separation performance of ion-exchange membranes for electrolytes in drainage nutrient solutions subjected to electrodialysis. Biosyst. Eng. 87, 89–97. https://doi.org/10.1016/j.biosystemseng.2003.09.005. Ligi, T., Truu, M., Truu, J., Nõlvak, H., Kaasik, A., Mitsch, W.J., Mander, Ü., 2014. Effects of soil chemical characteristics and water regime on denitrification genes (nirS, nirK, and nosZ) abundances in a created riverine wetland complex. Ecol. Eng. 72, 47–55. https://doi.org/10.1016/j.ecoleng.2013.07.015. Long, A., Heitman, J., Tobias, C., Philips, R., Song, B., 2013. Co-occurring anammox, denitrification, and codenitrification in agricultural soils. Appl. Environ. Microbiol. 79, 168–176. https://doi.org/10.1128/AEM.02520-12. McIsaac, G.F., David, M.B., Gertner, G.Z., Goolsby, D.A., 2001. Eutrophication: nitrate flux in the Mississippi River. Nature 414, 166–167. Park, J.B.K., Craggs, R.J., Sukias, J.P.S., 2009. Removal of nitrate and phosphorus from hydroponic wastewater using a hybrid denitrification filter (HDF). Bioresour. Technol. 100, 3175–3179. https://doi.org/10.1016/j.biortech.2009.01.036. Porter, M.D., 2011. Microbial community dynamics in denitrifying biofilters receiving agricultural drainage. (Dissertation of Master of Science). Department of Civil and

S. Fatehi-Pouladi et al. / Science of the Total Environment 655 (2019) 720–729 Environmental Engineering. University of Illinois at Urbana-Champaign, Champaign, IL, USA. Robertson, W.D., Cherry, J.A., 1995. In situ denitrification of septic-system nitrate using reactive porous media barriers: field trials. Groundwater 33, 99–111. https://doi.org/ 10.1111/j.1745-6584.1995.tb00266.x. Röling, W., van Breukelen, B., Braster, M., Goeltom, M., Groen, J., van Verseveld, H., 2000. Analysis of microbial communities in a landfill leachate polluted aquifer using a new method for anaerobic physiological profiling and 16s rDNA based fingerprinting. Microb. Ecol. 40, 177–188. https://doi.org/10.1007/s002480000033. Rozema, E.R., Gordon, R.J., 2014. Plant species for the removal of Na+ and Cl− from greenhouse nutrient solution. Hortscience 49, 1071–1075. Schipper, L.A., Cooper, A.B., Dyck, W.J., 1991. Mitigating nonpointsource nitrate pollution by riparian-zone denitrification. In: Bog'ardi, I., Kuzelka, R.D., Ennenga, W.G. (Eds.), Nitrate Contamination: Exposure, Consequence, and Control. Springer, Heidelberg, Germany, pp. 401–413 (Heidelberg, Germany). Schipper, L.A., Robertson, W.D., Gold, A.J., Jaynes, D.B., Cameron, S.C., 2010. Denitrifying bioreactors–an approach for reducing nitrate loads to receiving waters. Ecol. Eng. 36, 1532–1543. https://doi.org/10.1016/j.ecoleng.2010.04.008. Tanner, C.C., Sukias, J.P.S., Headley, T.R., Yates, C.R., Stott, R., 2012. Constructed wetlands and denitrifying bioreactors for on-site and decentralised wastewater treatment: comparison of five alternative configurations. Ecol. Eng. 42, 112–123. Warneke, S., Schipper, L.A., Matiasek, M.G., Scow, K.M., Cameron, S., Bruesewitz, D.A., McDonald, I.R., 2011. Nitrate removal, communities of denitrifiers and adverse effects

729

in different carbon substrates for use in denitrification beds. Water Res. 45, 5463–5475. https://doi.org/10.1016/j.watres.2011.08.007. Weber, K.P., 2016. Microbial community assessment in wetlands for water pollution control: past, present, and future outlook. Water 8. https://doi.org/10.3390/w8110503. Weber, K.P., Legge, R.L., 2009. One-dimensional metric for tracking bacterial community divergence using sole carbon source utilization patterns. J. Microbiol. Methods 79, 55–61. https://doi.org/10.1016/j.mimet.2009.07.020. Weber, K.P., Legge, R.L., 2010. Community-level physiological profiling. In: Cummings, S.P. (Ed.), Bioremediation: Methods and Protocols. Humana Press, Totowa, NJ,USA, pp. 263–281 https://doi.org/10.1007/978-1-60761-439-5_16. Weber, K.P., Grove, J.A., Gehder, M., Anderson, W.A., Legge, R.L., 2007. Data transformations in the analysis of community-level substrate utilization data from microplates. J. Microbiol. Methods 69, 461–469. https://doi.org/10.1016/j.mimet.2007.02.013. Zhai, X., Piwpuan, N., Arias, C.A., Headley, T., Brix, H., 2013. Can root exudates from emergent wetland plants fuel denitrification in subsurface flow constructed wetland systems? Ecol. Eng. 61, 555–563. Zhang, L., Lv, T., Zhang, Y., Button, M., Arias, C.A., Weber, K.P., Brix, H., Carvalho, P.N., 2018. Impacts of design configuration and plants on the functionality of the microbial community of mesocosm-scale constructed wetlands treating ibuprofen. Water Res. 131, 228–238.