Bioresource Technology 196 (2015) 490–499

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Enhancement and monitoring of pollutant removal in a constructedwetland by microbial electrochemical technology

http://dx.doi.org/10.1016/j.biortech.2015.07.1110960-8524/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +49 341 235 1357.E-mail address: carsten.vogt@ufz.de (C. Vogt).

Manman Wei a,b, Jana Rakoczy a, Carsten Vogt a,⇑, Falk Harnisch c, Reinhard Schumann a,Hans H. Richnow a

aDepartment of Isotope Biogeochemistry, Helmholtz Centre for Environmental Research-UFZ, Permoserstrabe 15, 04318 Leipzig, Germanyb Faculty of Natural Sciences, Hohenheim University, 70593 Stuttgart, GermanycDepartment of Environmental Microbiology, Helmholtz Centre for Environmental Research-UFZ, Permoserstraße 15, 04318 Leipzig, Germany

h i g h l i g h t s

� A bench scale HSSF-CW combined with MET was used to treat contaminated groundwater.� Significantly improved benzene, MTBE and NH4

+-N removal was obtained.� Benzene was oxidized at the anode, contributing to current generation.� MTBE and NH4

+-N removal were possibly enhanced due to eliminated inhibition effects.� The results imply a monitoring of benzene removal by current measurements.

a r t i c l e i n f o

Article history:Received 30 May 2015Received in revised form 27 July 2015Accepted 29 July 2015Available online 4 August 2015

Keywords:BenzeneMethyl-tert-butyl etherAmmoniumMETBioremediation

a b s t r a c t

A bench-scale constructed wetland combined with microbial electrochemical technology (MET-CW) wasrun for 400 days with groundwater contaminated with benzene, methyl-tert-butyl ether (MTBE), andammonium (NH4

+). Four vertically stacked anode modules were embedded into a sand bed and connectedwith a stainless steel cathode placed in an open water pond. In the zone of presence of anode modules,significantly more benzene and MTBE were removed in the MET-CW compared to the control CW with-out MET in the first 150 operation days. Benzene was identified as primary electron donor at the anode.Benzene removal and current densities were linearly correlated, implying the potential of the system forelectrochemically monitoring benzene biodegradation. Compound-specific isotope analysis (CSIA) indi-cated that benzene was initially activated by monohydroxylation forming intermediates which were sub-sequently oxidized accompanied by extracellular electron transfer, leading to current production. NH4

+

removal was not stimulated by MET.� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Microbial electrochemical technologies (METs) are a rapidlydeveloping technology platform at the nexus of microbiology andelectrochemistry (Schröder et al., 2015). The archetypes of METsare microbial fuel cells (MFCs), which can convert chemical energyof organic or inorganic compounds into electrical power. Recently,METs have been described to remove various pollutants from con-taminated waters such as organics (Zhang and He, 2012) as well asnitrogen and phosphorus containing compounds (Kelly and He,2014). Especially, the removal of nitrogen compounds (e.g. ammo-nia, nitrite, nitrate) in METs attracts increasing interest, providing a

competitive alternative to traditional wastewater treatment pro-cesses (Kelly and He, 2014). Recently, it was reported that micro-bial anodes in METs can effectively enhance the anaerobicbiodegradation of petroleum hydrocarbons (e.g. benzene, toluene),providing potential applications for the bioremediation of anaero-bic sediment or groundwater contaminated with petroleum hydro-carbons (Zhang et al., 2010; Wang et al., 2012; Wu et al., 2013). Inaddition to converting the chemical energy of pollutants into elec-tricity, METs were used as biosensors for on-line monitoring ofwastewater treatment or anaerobic digestion processes, such asMFC-based sensors for measuring the biological oxygen demand(BOD) (Di Lorenzo et al., 2009; Liu et al., 2011). Thus, microbialelectrochemical technologies possess the attractive potential toallow simultaneously removing pollutants, generating electricity,and monitoring the overall processes.

M. Wei et al. / Bioresource Technology 196 (2015) 490–499 491

Constructed wetlands (CWs) have been developed to treat pol-luted water for decades due to their low cost, easy operation andmaintenance (Imfeld et al., 2009; Garcia et al., 2010). In CWs, spa-tial gradients of dissolved oxygen (DO) concentrations and redoxconditions support the formation of ecological niches occupiedby several eco-physiologically different microorganisms, whichcan perform several different respiration and/or fermentation pro-cesses and generally show broad pollution removal capacities(Faulwetter et al., 2009; Imfeld et al., 2009). Spatially separatedoxic zones in the rhizosphere and anoxic zones in the deeper layersof a CW generate a potential difference which can be utilized tooperate a MFC, thereby providing an option for the enhancementof in situ pollutant removal while simultaneously generating elec-tricity. Systems efficiently combining MFCs and vertical flow (VF)or horizontal subsurface flow (HSSF) CWs have been previouslyreported (Yadav et al., 2012; Fang et al., 2013; Villasenor et al.,2013; Zhao et al., 2013). Villasenor et al. (2013) reported electricitygeneration during the treatment of synthetic wastewater with glu-cose as substrate in a lab-scale HSSF-CW by installing a horizontal,rectangular graphite cathode in the upper zone and an identicalanode in a deeper layer. However, the practical application is tech-nically challenging and more studies on the combined HSSF-CWand MET treating naturally polluted waters are essentially needed.

The Leuna site (Saxony-Anhalt, Germany) has been a refinerysite for more than 100 years, leading to massive groundwater con-tamination with gasoline compounds, especially benzene andmethyl-tert-butyl ether (MTBE). Additionally, the groundwater ishighly contaminated with ammonium (NH4

+) deriving from a for-mer ammonia production site. Therefore, the Leuna site was cho-sen as the reference site for testing the bioremediation ofpetroleum hydrocarbons and NH4

+ contaminated groundwater withseveral different pilot-scale HSSF-CWs (Seeger et al., 2011; vanAfferden et al., 2011; Chen et al., 2012). Recently, Wei et al.(2015) described a small-scale MFC reactor with an aerated cath-ode by which benzene and NH4

+-N contaminated groundwater fromthis site was successfully treated under controlled laboratory con-ditions; benzene biodegradation at the anode was coupled withelectricity generation while ammonium was oxidized at the cath-ode. Although METs and CWs technologies have been separatelyapplied for the effective remediation of contaminated groundwaterat the Leuna site, it is worthwhile to develop a combined MET andCW system considering the possibility for exploiting benefits ofboth METs and CWs.

In this study, an integrated MET-CWwas established by embed-ding four anode modules into the sand bed and connecting it to acathode placed in the open pond inside a bench-scale HSSF-CW;the system was used to treat original Leuna groundwater contam-inated with benzene, MTBE and NH4

+. The main objective of thestudy was to evaluate whether MET can be used to improve thebiodegradation of the main contaminants. Furthermore, the perfor-mances of anode modules located at the different depths werecompared in order to determine favorable depths for anodicreactions.

2. Methods

2.1. Reactor construction and set-up

The MET reactor was made of acrylic material and comprisedfour individual anode modules (each 10.5 � 4.2 � 3.3 cm, 53 cmheight in total) as shown in Fig. 1A. Each module contained 27holes (diameter 1 cm) covered by nylon gauze, allowing waterflow-through. The anode consisted of six parallel graphite plates(9 � 3.5 � 0.4 cm), connected each other by a graphite rod with adiameter of 0.5 cm, leading to a total specific surface area of

0.0438 m2 per module. The four anode modules were connectedwith a stainless steel cathode (304 SS, 13.5 � 3.5 � 1 cm) acrossthe constant external resistances of 200 O for operation and vary-ing resistances for polarization curves, respectively.

2.2. Setup and operation of the MET-CW system

A general scheme of the MET-CW system and detailed informa-tion on the experimental design are provided in Fig. 1B. The hori-zontal subsurface flow wetland consisted of a stainless steel tank(201 � 60 � 5 cm) equipped with a 1 cm thick glass panel on thefront side, filled to an average depth of 52 cm with quartz sand(grain size = 0.68–0.80 mm) and planted with common rush(Juncus effuses, L.). The glass panel was covered by a removablelight-tight board to avoid light effects. At 150 cm distance fromthe inflow, a 50 cm long open water pond was created, remainingin direct contact with the atmosphere. The sand bed was separatedfrom the open water pond using a stainless steel plate, which con-tained several holes (diameter 1 cm) covered by nylon gauze,allowing the water to flow through. At the inflow, four inlet pipeswere located 4, 12, 20, and 36 cm above the bottom to facilitate auniform distribution of groundwater through the whole wetlandbody. Due to the presence of a pore at the outflow part, the waterlevel was kept at 48 cm depth from the bottom. Anoxic groundwa-ter was taken from a contaminated site close to the Leuna refinery(Saxonia, Anhalt, Germany) and kept in a 50 L stainless steel tankunder constant dinitrogen pressure (1.5 bar); this water was con-tinuously pumped into the MET-CW at a constant flow rate(0.3 mL min�1) by a peristaltic pump (Ismatec REGLO Analog MS-2/6, Wertheim, Germany), yielding an inflow of 1.73 liters perday and resulting in a hydraulic retention time (HRT) of approxi-mately 15 days. The composition of the contaminated groundwateris shown in Supplementary Table S1. The system was equippedwith a cooling system maintaining the temperature of the tankand the wetland constantly at 14 ± 2 �C. Four anode modules wereembedded into the sand bed of the CW located between the sam-pling ports at the flow distance of 94 cm and 139 cm. The modulesides containing the holes were placed in the direction of the waterflow, allowing water to pass through. The cathode was installed atthe surface of the open water body with around 80 cm distancefrom the anode modules. A control CW containing no MET wassimilarly constructed and operated; the mechanisms of benzenebiodegradation and hydrogeological parameters in this systemhave been previously reported by Rakoczy et al. (2011).

2.3. Electrochemical measurements and calculations

The voltage (V) across the resistor (R) was recorded at 20 minintervals using a Paperless Recorder (JUMO LOGOSCREEN 500 cf,Fulda, Germany). The current (I) was calculated by Ohm’s law(I = V/R). The current and power densities were normalized bythe total surface area per anode module (A). The coulombic effi-ciency (CE) was calculated as the ratio of total electrons recoveredas charge to maximum possible electrons released by total sub-strate removal. Polarization and power density curves wereobtained by varying the resistance in a range from 10.000 to100 O using a multimeter (Metrix MTX 3282, Paris, France).Therefore, the reactor was first disconnected for 5 h to record astable open circuit voltage (OCV). Data for each resistance adjust-ment were recorded for at least 30 min interval after the voltagechange was less than 2 mV in 1 min. The measurement was carriedout in triplicates; a constant resistance of 200 O was reconnectedovernight between every replicate measurement. The internalresistance was calculated based on the slope of the linear regionof the polarization curve.

Fig. 1. Schematic diagram of the MET-CW used in this study.

492 M. Wei et al. / Bioresource Technology 196 (2015) 490–499

2.4. Sampling procedure

Twelve sampling ports were installed for collecting water sam-ples across the bed (Fig. 1). Four vertical groups across the sand bedlocated at 6, 49, 94 and 139 cm distance from the inflow, each con-sisting of three sampling ports at 24, 36 and 48 cm depth from thetop, were systematically investigated. Water samples from theinflow, outflow and the 12 sampling ports were collected at differ-ent operation times using a gas-tight syringe through a valve aspreviously described (Rakoczy et al., 2011). The sampling timesare indicated in Figs. 2 and 3 and S2. In order to analyze differentbiogeochemical parameters, each pore water sample consisted ofsix subsamples which were stored in completely filled sterile20 mL vials. Two subsamples for benzene and MTBE measure-ments were immediately fixed using sodium hydroxide to inhibitmicrobial activity. One subsample was adjusted to a pH of 2 withHCl (36%) in order to measure ferrous and total ionic iron. The

other water samples were stored at 4 �C until analysis. Each samplewas analyzed at least three times (see Section 2.5), and the meanwas calculated. The outflow volumes were measured in compar-ison to the inflow volumes in order to determine water loss, allow-ing the calculation of pollutant load removal.

2.5. Chemical analysis and calculations

Water samples were analyzed for the concentrations of ben-zene, MTBE, ammonium (NH4

+), nitrite (NO2�), nitrate (NO3

�), sulfate(SO4

2�), phosphate (PO43�), chloride (Cl�), ferrous iron (Fe2+), and

total ionic iron. Briefly, benzene and MTBE were analyzed using agas chromatograph equipped with a flame ionization detector(Varian CP-3800 GC, Palo Alto, CA) as described elsewhere(Fischer et al., 2008; Rosell et al., 2010). The concentrations ofNH4

+, NO2� and NO3

� were analyzed colorimetrically as describedbefore (Bollmann et al., 2011). The pH was monitored by a pH

Fig. 2. Removal efficiencies of benzene, MTBE, and NH4+ in the MET-CW and the

control CWwithin 400 days, calculated based on the averaged mass loads for 24 cm,36 cm and 48 cm depth at the sampling point after 139 cm flow path (see Fig. 1B).*Represents significant differences (p < 0.05) between the MET-CW and the controlCW. Original concentration data in the control CW at days 231, 293 and 345 werealready published by Rakoczy et al. (2011).

M. Wei et al. / Bioresource Technology 196 (2015) 490–499 493

meter (Knick, Berlin, Germany). The DO concentration was deter-mined in situ by an optical trace sensor system (PreSens sensorspot PSt6 and FIBOX-3 minisensor oxygen meter, 170Regensburg, Germany). The oxygen sensor spots were attached tothe inner side of the glass panel, distributed corresponding to thesampling ports across the sand bed. The redox potential (Eh) wasmeasured with a pH/mV/Temp meter (Jenco Electronics 6230N,San Diego, USA).

Putative water losses through plant transpiration and evapora-tion from the surface (evapotranspiration) were taken intoaccount in order to eliminate the effect of evapotranspiration onpollutant concentrations (Hijosa-Valsero et al., 2010). It was

assumed that the water loss along the flow path followed a linearchange. Therefore, the mass of pollutants was calculated asfollows:

Mij ¼ Cij � Q inf �ðQ inf � Q effÞ

L� Li

� �ð1Þ

where Mij is the pollutant mass at distance i cm and depth j cm(mg d�1); Cij is the pollutant concentration at distance i cm anddepth j cm (mg L�1); Qinf is the influent flow rate (L d�1); Qeff isthe effluent flow rate (L d�1); Li is the distance from inflow to thesampling point i (cm); and L is the entire length of the CW (cm).

The pollutant removal efficiency was calculated based on thedifferences of pollutant mass between the influent and the sam-pling points using the following equation:

Removal efficiency ð%Þ ¼ 1� Mij

Minf

� �o100 ð2Þ

where Minf is the influent mass load (mg d�1) and Mij is the massload of the sampling point i (mg d�1).

2.6. Compound-specific stable isotope analysis

Samples from the inflow and sampling ports (see Section 2.4)were collected at operation day 95 and extracted with pentaneas previously described (Wei et al., 2015). The carbon and hydro-gen stable isotope compositions were determined using a gas chromatograph-combustion-isotope ratio mass spectrometer system(GC-IRMS). Measurements and calculations were performed asdescribed by Rakoczy et al. (2013). The benzene degradation path-ways were analyzed by comparing measured isotope compositionshifts with published isotope enrichment factors (e) in a two-dimensional isotope plot (Fischer et al., 2008).

2.7. Statistical analysis

Statistical analyses were performed using the SPSS 22.0 package(Chicago, IL, USA). The normality and homogeneity were assessedwith a Shapiro–Wilk W test and a Levene test, respectively. If thedata were normally distributed, one-way ANOVA was used to ana-lyze differences in pollutant removal efficiencies between theMET-CW and control CW. The Tukey0s post-hoc test was used tofurther evaluate the difference between the different depths withinthe same flow paths when significant differences were found. Non-parametric Mann–Whitney U-tests were performed instead ofANOVA tests, as a normal distribution could not be assumed.Differences were considered to be statistically significant ifp < 0.05. The linear relationship between pollutant loading removal(mass loading difference between 94 and 139 cm) and current den-sities were analyzed by a regression analysis.

3. Results and discussion

3.1. Removal efficiencies of benzene, MTBE and NH4+-N

The major pollutants in the influent groundwater were ben-zene, MTBE and NH4

+ (Table S1), of which corresponding concentra-tions were up to 12, 3, and 50 mg L�1, respectively. The removalefficiencies of these compounds within the MET-CW and controlCW during the whole experimental period are shown in Fig. 2.The depth-dependent spatial variations of removal efficienciesalong the flow path were also determined in order to clarify therole of MET upon pollutant degradation and to detect the mostsuitable locations of the anode modules; this is exemplarily shownin Fig. 3 for the MET-CW at operation day 95 and the control CW atday 97, respectively. Pollutant removal efficiencies reached steady

Fig. 3. Pollutant removal efficiencies along the flow path in the MET-CW at the 95th day and the control CW at the 97th day. Electricity was effectively generated in the MET-CW (see Fig. 4). *Represents significant differences (p < 0.05) between the MET-CW and the control CW at the same sampling sites. a,b,cThe different letters indicate significantdifferences (p < 0.05) between the different depths at the same flow paths.

494 M. Wei et al. / Bioresource Technology 196 (2015) 490–499

state in both the MET-CW and control CW after around 150 days,while current generation of the MET-CW started to decrease andfinally slowly disappeared. Thus, the first 150 days of operationwere termed phase I, and the subsequent experimental time wastermed phase II.

In phase I, the overall removal efficiencies of the three investi-gated pollutants gradually increased in both the MET-CW and con-trol CW, indicating the attachment and growth of degraders as itusually occurs in wetland systems (Saeed and Sun, 2012). In thefirst 40 operation days, the removal efficiencies of benzene,MTBE and NH4

+-N were similar (p > 0.05) in both systems (Fig. 2).From day 40 to 150, significant differences (p < 0.05) wereobserved between the MET-CW and control CW for benzene andMTBE removal. In particular, at day 95 and 125, benzene andMTBE were nearly completely removed in the MET-CW, whereassuch high removal efficiencies were obtained in the control CWonly after 150 days (Fig. 2). In phase II (i.e. after 150 days), benzeneand MTBE were almost completely removed in the MET-CW andcontrol CW. Notably, higher removal rates of benzene compared

to MTBE have been reported in pilot-scale CWs at the field siteLeuna (Jechalke et al., 2010; Seeger et al., 2011; Chen et al.,2012). Thus, the observed high removal efficiencies for both ben-zene and MTBE in phase I indicate that MTBE degradation wasnot affected by the presence of benzene. BTEX compounds (ben-zene, ethylbenzene, toluene and xylene) have been reported toinhibit MTBE degraders (Deeb et al., 2001), which usually hampersthe efficient treatment of waters contaminated with MTBE andpetroleum compounds (van Afferden et al., 2011).

Fig. 3 shows the removal efficiencies of benzene, MTBE andNH4

+-N at different sampling points and depths in the MET-CW atday 95 and the control CW at day 97, respectively. Overall, highestremoval efficiencies of benzene and MTBE were observed in theupper layer (24 cm), probably due to a better availability of oxygenreflected by the higher concentrations of DO in this layer(Supplementary Table S2). Compared to the control CW, slightlymore MTBE was removed in the upper layer of all four distancesin the MET-CW; for benzene, this effect was observed only in theupper layer at 94 cm and 139 cm distance (Fig. 3). A hot zone of

M. Wei et al. / Bioresource Technology 196 (2015) 490–499 495

degradation was detectable in the MET-CW between 94 cm and139 cm distance; here, the removal efficiencies of benzene andMTBE in the MET-CW were significantly higher in all three investi-gated depths (24 cm, 36 cm, 48 cm) compared to the control CW(p < 0.05) (Fig. 3). Notably, the anode modules of MET were placedin this zone (Fig. 1), indicating that this effect was caused by METoperation. Differences of MTBE and benzene removal efficienciesbetween the MET-CW and control CW were much less pronouncedat 6 cm, 49 cm and 94 cm distance from the inflow (Fig. 3). Thephysico-chemical parameters (pH, DO, Eh, Fe2+ and SO4

2�) wererather similar in the MET-CW and control CW at all depths alongthe whole flow path (Table S2), indicating that the physico-chemical conditions were generally not changed by the insertionof the anode. In the zone where the MET anodes were placed(between the sampling points 94 cm and 139 cm), significantlymore benzene and MTBE were removed in the deeper layers (36and 48 cm depth) where the anode modules 3 and 4 were placedand where the actual DO concentrations were relatively low(<0.5 mg L�1). Due to the overall higher removal rates in the upperlayers of 24 cm depth (Figs. 2 and 3), a large fraction of benzene(88%) and MTBE (30%) was already degraded at 94 cm distance,and only a relatively low percentage of them reached the upperanode modules, thereby leading to much less benzene and MTBEremoval at 24 cm depth.

In contrast to the results for benzene and MTBE, NH4+ removal

efficiencies were similar between the MET-CW and control CW inphase I (p > 0.05). However, the mean NH4

+ removal efficiency inphase II was significantly higher in the MET-CW (almost 100%)than that observed in the control CW (62%) (p < 0.05) (Fig. 2).NO3

� and NO2� accumulated in the upper layers (24 cm) of the

MET-CW and control CW (Fig. S2), demonstrating that nitrificationwas a major removal process. In contrast, NO3

� and NO2� were not

detected in the lower layers of 36 and 48 cm depth, althoughNH4

+ concentrations decreased. This zonation was formed probablydue to the levels of DO as average DO values of 0.75 and0.70 mg L�1 were detected in the upper layers of MET-CW and con-trol CW, respectively, whereas lower average values varyingbetween 0.04 and 0.21 mg L�1 were measured in the lower layersof both CWs (Supplementary Table S2). DO concentrations of lessthan 0.5 mg L�1 are known to promote denitrification or anaerobicammonium oxidation and thereby consuming NO3

� and/or NO2�

produced from nitrification (Saeed and Sun, 2012). The transforma-tion processes of inorganic nitrogen in the MET-CW will be moredeeply described in a separate manuscript (Wei et al., inpreparation).

It is known that the availability of substrates and their oxida-tion activities at the anode are of great importance for electricitygeneration in METs. In this study, benzene and MTBE were themain organic substrates in the contaminated groundwater andhence expected to be potential electron donors for the anodic reac-tion. Our data indicate that the removal of benzene and MTBE inthe first 150 operation days was slightly stimulated by the pres-ence of anode modules in the MET-CW. After 150 days, benzeneand MTBE concentrations in the MET-CW already stronglydecreased in the front part of the bed due to the enhancedbiodegradation by mature microbial communities, thereby cuttingoff the anode modules from its fuels, e.g. benzene or MTBE.

Compared with the control CW, higher NH4+ removal rates in the

MET-CW were detected only after 150 days operation. The pres-ence of organic compounds and their metabolic intermediateswere described to inhibit the nitrification process in general(Saeed and Sun, 2012), as organo-heterotrophic degraders poten-tially outcompete nitrifiers for the utilization of oxygen and inor-ganic nitrogen (Ma et al., 2013), thereby resulting in lownitrification rates under oxygen-limited conditions. The presenceof benzene and MTBE probably inhibited the removal of NH4

+, as

also previously observed in field studies at the Leuna site(Jechalke et al., 2011; Seeger et al., 2011). However, the rapid con-sumption of benzene and MTBE after 150 days probably lead to abetter availability of oxygen for nitrifiers, thereby promoting theNH4

+ removal in the MET-CW.

3.2. Electricity recovery in the MET-CW

3.2.1. The electric energy generationThe current densities generated from the four anode modules of

the MET-CW were recorded during the whole operation as shownin Fig. 4A. The highest current output was observed in phase I:here, average current densities of 0.86 ± 0.74 and1.02 ± 0.79 mAm�2 were obtained in modules 3 and 4, with peakvalues of 3.39 and 5.46 mAm�2, respectively. In contrast, modules1 and 2 produced very low current densities during the wholeoperation, which can be ascribed to the presence of oxygen. Inphase II, the current densities of modules 3 and 4 decreased signif-icantly; after 280 days, the current generation ceased completely.Low or no current generation in phase II can be explained by thefact that the more efficient benzene and MTBE degradationoccurred in the front of the bed after the system had been fullydeveloped (Fig. 2), resulting in less available substrates for theanode modules. The obtained maximum current densities in mod-ules 3 and 4 are too low to make the MET practically competitivefor electricity recovery. However, the relationship between theavailability of organic substrates and current generation impliesthe possible application of this MET as a sensor for on-line moni-toring pollutant degradation processes in situ, as discussed in moredetail in Section 3.3. Huge fluctuations of the current density wereprobably due to slight variations of the experimental conditions,such as influent pollutant concentrations, pH, or redox potentials(Table S2). Fluctuating conditions have been also reported for theintegrated HSSF-CW and MFC system studied by Villasenor et al.(2013).

3.2.2. Polarization and power density curvesThe electrochemical performance of the MET-CWwas evaluated

by determining polarization and power density curves (Fig. 4B),which were obtained at day 100 in phase I. Modules 3 and 4showed maximal power densities of 0.86 mWm�2 at a currentdensity of 3.62 mAm�2, and 1.74 mWm�2 at a current density of5.75 mA m�2, respectively. In contrast, the power densities of mod-ules 1 and 2 were extremely low with maximum values of 0.01 and0.03 mWm�2 respectively (corresponding to the observed extre-mely low current densities, Fig. 4A), which was attributed to thenoise current and thus considered to be negligible. These resultssuggest that the anodes located in the deeper layer (modules 3and 4) supported solely the electricity generation. The very lowcurrent generations of modules 1 and 2 are possibly due to higherDO concentrations (Table S2) disturbing the anodic reaction of theMET. Oxygen in the upper layers – stemming from atmosphericdiffusion – will outcompete the anode as electron acceptor. Thisresult was also consistent with the removed masses of benzeneand MTBE as described in Section 3.1.

In recently described planted MET-CW systems with vertical orhorizontal subsurface flow (Fang et al., 2013; Villasenor et al.,2013; Zhao et al., 2013), the cathodes were installed in the moreaerobic upper layer. In the present study, the cathode was locatedin the open water pond, avoiding the recirculation of pollutantsfrom the anodic compartment to the cathodic compartment; inaddition, an open water pond is more effective for storing oxygen,thereby eliminating the negative effect of the residual organic mat-ter on the maintenance of aerobic conditions at the cathode underhigh organic loading rates as observed by Villasenor et al. (2013).

Fig. 4. Electricity generation performance of the anode modules in the MET-CW. (A) current generation during the whole operation period; (B) polarization and powerdensity curves. Average values and standard deviation were calculated from measurements of three replicates (details see Section 2.3).

496 M. Wei et al. / Bioresource Technology 196 (2015) 490–499

The maximum power density of 1.74 mWm�2 obtained in thiswork is relatively low compared to previously published studies(Yadav et al., 2012; Villasenor et al., 2013; Zhao et al., 2013), inwhich values in the range between 10 and 50 mWm�2 wereobserved (Table S4). The relatively low power densities obtainedin this study might be due to lower amounts of available sub-strates, differences in the electrochemical properties of substratesor other operation and construction parameters (e.g., the magni-tude of the internal resistance). However, they are comparablewith values obtained in the MFC systems treating waters contam-inated with petroleum hydrocarbons, e.g. benzene (Wang et al.,2012; Rakoczy et al., 2013; Wu et al., 2013), where power densitiesfrom 0.85 to 2.1 mWm�2 were reported. Recently, a maximumpower density of 316 mWm�3 NAC (net anodic compartment)and a coulombic efficiency of 14% was obtained in a lab-scaleMFC treating the same groundwater contaminated with benzene,MTBE and NH4

+-N as used in the present study, in which graphitegranules were used as electrode materials (Wei et al., 2015). Inour study, coulombic efficiencies were calculated based on ben-zene removal since benzene was identified as primary anodic sub-strate (see Section 3.3). The average coulombic efficiencies of3.69 ± 0.81% and 4.89 ± 0.60% were obtained for the modules 3and 4 in phase I, which are comparable with values reported forother CW-MFC systems (Table S4).

Based on the slopes of the polarization curves, internal resis-tances of 1588 ± 83 and 1253 ± 41 O were obtained for modules

3 and 4, respectively. It is generally known that the power densityreaches its peak value when the external resistance is equal to itsinternal resistance (Harnisch and Schröder, 2010). According tothe power density curves, the maximum power densities wereobtained at the external resistance of 1500 and 1200 O for module3 and 4, consistent with the values calculated from the slopes ofthe polarization curves. The high internal resistances were proba-bly caused by the relative long distance (�80 cm) between theanode modules and the cathode in the MET-CW, leading to adecrease of the electricity generation performance. However, thosevalues are still in an acceptable range; similar values were reportedin our previous study using a ‘classical’ MFC system (Wei et al.,2015).

3.3. Correlations between pollutant removal and electricity generation

In order to further clarify the relationship of pollutant removaland electricity generation in the MET-CW, a regression analysisbetween pollutant loading removal by the anode modules and gen-erated current density was performed. When integrating the threepollutants by multiple-regression analysis, benzene degradationwas identified to be the primary correlation factor (p < 0.05),explaining 99% and 100% of the variation in current generationfor the modules 3 and 4, respectively; as shown in Fig. 5, benzeneloading removal rates and current densities were strongly corre-lated (R = 0.99, p < 0.05). The steeper slope for module 4 indicated

M. Wei et al. / Bioresource Technology 196 (2015) 490–499 497

that this module generated current more efficiently; correspond-ingly, the higher current generation was also obtained in module4 (Fig. 4). In contrast, removal of MTBE and NH4

+ was not signifi-cantly associated with current densities in the modules 3 and 4(p > 0.05), implying no direct link between MTBE and NH4

+ removaland current generation in the MET-CW. Thus, current was gener-ated mainly by oxidation of benzene at the anode.

The strong linear correlation between current generation andbenzene removal indicates the possibility for monitoring ben-zene degradation using a MET based biosensor. In order to jus-tify and evaluate the potential for the practical application, theoperational parameters, such as reproducibility, stability anddetection limit, need to be clarified in future studies.Nevertheless, our results indicate a promising technique of mon-itoring the removal of organic pollutants in constructed wetlandson-line and in situ.

3.4. Electron donors for the anode modules

The redox potentials for the oxidation of benzene (HCO3�/C6H6,

E00 = �0.28 V) and MTBE (HCO3�/C5H12O, E00 = �0.31 V) are in the

same range, thus both compounds may react similarly as anodicelectron donors under identical condition (Table S3). However,our data demonstrate that bioelectrochemical benzene oxidationis more efficient, which is in accordance with the generally betterbiodegradability of benzene in co-occurrence with MTBE (Deebet al., 2001). MTBE degradation is usually hindered by the presenceof BTEX due to competitive inhibition or low growth rates ofMTBE-utilizing microorganisms. Furthermore, the influent concen-tration of MTBE (�3 mg mL�1) was four-fold lower than benzene(�12 mg mL�1), resulting in a higher availability of benzene forthe anode modules. Thus, benzene was identified as the primaryelectron donor rather than MTBE in this case. However, the rapidconsumption of benzene stimulated by the anode modules elimi-nates potential inhibitory effects on MTBE degradation, andthereby promoting MTBE removal in the MET-CW in the first150 days.

Although ammonium removal from wastewater using electro-chemical technologies has been previously reported, there is nounanimous agreement whether ammonium can be used as a directsubstrate for electricity generation in METs. He et al. (2009)showed that electricity production is associated with ammoniumoxidation in a rotating-cathode MFC, in which current generation

Fig. 5. Linear relationship between benzene loading removal by the anode modulesand current generation in the MET-CW. The pollutant loading removal wascalculated as the difference in pollutant loads between the sampling points at94 cm and 139 cm.

resulted from heterotrophic exoelectrogenesis sustained bynitrifier-produced organic substrates rather than direct ammo-nium oxidation. Furthermore, You et al. (2009) reported that addi-tional protons produced by ammonium oxidation can acceleratethe cathodic reaction (e.g. oxygen reduction), thereby contributingto electricity generation by reducing the ohmic resistance andmaintaining the pH balance. However, intermediates (hydroxy-lamine and NO2

�) of ammonium oxidation can release electronsin the view of thermodynamic properties and may possibly serveas anodic substrate to generate current (Chen et al., 2014); thereduction of the end product (NO3

�) at the cathode can also resultin electrochemical denitrification. In this study, NH4

+ oxidationwas not directly linked with electricity generation in the MET-CW in the presence of benzene and MTBE as co-contaminants.Nitrification reactions are characterized by more positive redoxpotentials and generate less available energy compared to benzeneand MTBE oxidation reactions (Table S3), suggesting that NH4

+ is aless favorable MFC substrate. Correspondingly, Wei et al. (2015)reported that NH4

+ oxidation did not contribute to electricity gener-ation in a MFC with aerated cathode. The higher removal of NH4

+ inphase II can be attributed to the elimination of the inhibitoryeffects caused by benzene and MTBE due to their rapidconsumption.

3.5. Implications for the mechanism of benzene activation

Compound-specific isotope analysis (CSIA) was performed inorder to identify the initial activation mechanisms for benzenedegradation in the MET-CW. Application of CSIA to characterizebenzene activation mechanisms has been successfully demon-strated in laboratory and field studies (Fischer et al., 2008;Rakoczy et al., 2011, 2013). A significant shift in the carbon andhydrogen isotope signature was observed with Dd13C up to+2.7‰ and Dd2H up to +15‰ (Fig. S3). By comparing the valuesof two-dimensional CSIA with previously published values for aer-obic and anaerobic benzene activation mechanisms, the data fromour study mainly matched with those indicative for pathways ini-tiated by benzene monohydroxylation (Fig. 6). The slope of the lin-ear regression between hydrogen and carbon isotopediscrimination (K = Dd2H/Dd13C) obtained in the MET-CW wasK = 7 ± 2 (R2 = 0.71). It has been predicted that the initial activationstep of monohydroxylation reaction leads to K-values ranging

Fig. 6. Two-dimensional isotope plot of Dd13C versus Dd2H values of benzenemeasured from pore water samples of the MET-CW at the 95th day. Data were fromthe depths of 36 cm and 48 cm. Values of Dd were calculated by subtracting themeasured isotopic value from the initial isotopic value determined at the influent.

498 M. Wei et al. / Bioresource Technology 196 (2015) 490–499

between 3 and 11 depending on the reaction mechanism ofmonooxygenase (Fischer et al., 2008), which can be distinguishedsignificantly from pathways initiated by (i) benzene dihydroxyla-tion, a reaction which is usually not linked to a significant hydro-gen isotope effect, and (ii) anaerobic benzene degradation, whichis usually characterized by the larger K values (K > 13). Benzenemonohydroxylation as initial activation mechanism of benzenedegradation was already reported by Rakoczy et al. (2011) forthe control-CW, and also identified in a MFC treating benzeneand sulfide-contaminated groundwater (Rakoczy et al., 2013).However, we cannot exclude that benzene was actually anaerobi-cally activated by a mechanism producing similar carbon andhydrogen isotope fractionation as observed for monohydroxyla-tion, or that the isotope signatures were produced by a mixing ofaerobic and anaerobic activation mechanisms due to the coexis-tence of oxic and anoxic zones in the CW. Due to the detection ofat least small amounts of DO at each sampling location duringthe whole experimental time (Table S2), strictly anoxic processesmay not have played a significant role for benzene degradation.In any case, the degradation of benzene and/or its intermediatesaccelerated electricity generation by transferring the released elec-trons to the anode in the MET-CW.

4. Conclusions

This work demonstrates that benzene, MTBE and NH4+ removal

in a constructed wetland can be improved by combination withmicrobial electrochemical technology. The enhanced benzeneremoval was linked to its direct use as electron donor by the anodemodules, whereas efficient removal of MTBE and NH4

+ was proba-bly attributed to the elimination of inhibition effects by the co-contaminant benzene. The linear relationship between currentdensity and benzene removal implies the potential biosensorapplication for monitoring benzene oxidation processes in situ.

Acknowledgements

This work was financially supported by the Helmholtz Centrefor Environmental Research-UFZ in the scope of the SAFIRA IIResearch Project (Revitalization of Contaminated Land andGroundwater at Megasites, subproject ‘‘Compartment Transfer-CoTra”). F.H. acknowledges the support by the BMBF (ResearchAward ‘‘Next generation biotechnological processes –Biotechnology 2020+”) and the Helmholtz-Association (YoungInvestigators Group).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biortech.2015.07.111.

References

Bollmann, A., French, E., Laanbroek, J.L., 2011. Isolation, cultivation, andcharacterization of ammonia-oxidizing bacteria and archaea adapted to lowammonium concentrations. In: Abelson, J., Melvin, I.S., Colowick, S.P., Nathan, A.K. (Eds.), Methods in Enzymology. Elsevier, Oxford, USA, pp. 55–88, Vol.Research on nitrification and related processes, part A.

Chen, H., Zheng, P., Zhang, J., Xie, Z., Ji, J., Ghulam, A., 2014. Substrates and pathwayof electricity generation in a nitrification-based microbial fuel cell. Bioresour.Technol. 161, 208–214.

Chen, Z., Kuschk, P., Reiche, N., Borsdorf, H., Kastner, M., Koser, H., 2012.Comparative evaluation of pilot scale horizontal subsurface-flow constructedwetlands and plant root mats for treating groundwater contaminated withbenzene and MTBE. J. Hazard. Mater. 209–210, 510–515.

Deeb, R.A., Hu, H.Y., Hanson, J.R., Scow, K.M., Alvarez-Cohen, L., 2001. Substrateinteractions in BTEX and MTBE mixtures by an MTBE-degrading isolate.Environ. Sci. Technol. 35, 312–317.

Di Lorenzo, M., Curtis, T.P., Head, I.M., Scott, K., 2009. A single-chamber microbialfuel cell as a biosensor for wastewaters. Water Res. 43 (13), 3145–3154.

Fang, Z., Song, H.L., Cang, N., Li, X.N., 2013. Performance of microbial fuel cellcoupled constructed wetland system for decolorization of azo dye andbioelectricity generation. Bioresour. Technol. 144, 165–171.

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 oftreatment wetlands: a review. Ecol. Eng. 35 (6), 987–1004.

Fischer, A., Herklotz, I., Herrmann, S., Thullner, M., Weelink, S.A., Stams, A.J.,Schlomann, M., Richnow, H.H., Vogt, C., 2008. Combined carbon and hydrogenisotope fractionation investigations for elucidating benzene biodegradationpathways. Environ. Sci. Technol. 42 (12), 4356–4363.

Garcia, J., Rousseau, D.P.L., MoratÓ, J., Lesage, E.L.S., Matamoros, V., Bayona, J.M.,2010. Contaminant removal processes in subsurface-flow constructedwetlands: a review. Crit. Rev. Environ. Sci. Technol. 40 (7), 561–661.

Harnisch, F., Schröder, U., 2010. From MFC to MXC: chemical and biologicalcathodes and their potential for microbial bioelectrochemical systems. Chem.Soc. Rev. 39 (11), 4433–4448.

He, Z., Kan, J., Wang, Y., Huang, Y., Mansfeld, F., Nealson, K.H., 2009. Electricityproduction coupled to ammonium in a microbial fuel cell. Environ. Sci. Technol.43 (9), 3391–3397.

Hijosa-Valsero, M., Matamoros, V., Sidrach-Cardona, R., Martin-Villacorta, J.,Becares, E., Bayona, J.M., 2010. Comprehensive assessment of the designconfiguration of constructed wetlands for the removal of pharmaceuticals andpersonal care products from urban wastewaters. Water Res. 44 (12), 3669–3678.

Imfeld, G., Braeckevelt, M., Kuschk, P., Richnow, H.H., 2009. Monitoring andassessing processes of organic chemicals removal in constructed wetlands.Chemosphere 74, 349–362.

Jechalke, S., Rosell, M., Vogt, C., Richnow, H.H., 2011. Inhibition of nitrification bylow oxygen concentrations in an aerated treatment pond system with biofilmpromoting mats. Water Environ. Res. 83 (7), 622–626.

Jechalke, S., Vogt, C., Reiche, N., Franchini, A.G., Borsdorf, H., Neu, T.R., Richnow, H.H., 2010. Aerated treatment pond technology with biofilm promoting mats forthe bioremediation of benzene, MTBE and ammonium contaminatedgroundwater. Water Res. 44 (6), 1785–1796.

Kelly, P.T., He, Z., 2014. Nutrients removal and recovery in bioelectrochemicalsystems: a review. Bioresour. Technol. 153, 351–360.

Liu, Z., Liu, J., Zhang, S., Xing, X.H., Su, Z., 2011. Microbial fuel cell based biosensorfor in situ monitoring of anaerobic digestion process. Bioresour. Technol. 102(22), 10221–10229.

Ma, J., Wang, Z., Zhu, C., Liu, S., Wang, Q., Wu, Z., 2013. Analysis of nitrificationefficiency and microbial community in a membrane bioreactor fed with lowCOD/N-ratio wastewater. PLoS One 8 (5), e63059.

Rakoczy, J., Feisthauer, S., Wasmund, K., Bombach, P., Neu, T.R., Vogt, C., Richnow, H.H., 2013. Benzene and sulfide removal from groundwater treated in a microbialfuel cell. Biotechnol. Bioeng. 110 (12), 3104–3113.

Rakoczy, J., Remy, B., Vogt, C., Richnow, H.H., 2011. A bench-scale constructedwetland as a model to characterize benzene biodegradation processes infreshwater wetlands. Environ. Sci. Technol. 45 (23), 10036–10044.

Rosell, M., Finsterbusch, S., Jechalke, S., Hubschmann, T., Vogt, C., Richnow, H.H.,2010. Evaluation of the effects of low oxygen concentration on stable isotopefractionation during aerobic MTBE biodegradation. Environ. Sci. Technol. 44 (1),309–315.

Saeed, T., Sun, G., 2012. A review on nitrogen and organics removal mechanisms insubsurface flow constructed wetlands: dependency on environmentalparameters, operating conditions and supporting media. J. Environ. Manage.112, 429–448.

Schröder, U., Harnisch, F., Angenent, L.T., 2015. Microbial electrochemistry andtechnology: terminology and classification. Energy Environ. Sci. 8, 513–519.

Seeger, E.M., Kuschk, P., Fazekas, H., Grathwohl, P., Kaestner, M., 2011.Bioremediation of benzene-, MTBE- and ammonia-contaminatedgroundwater with pilot-scale constructed wetlands. Environ. Pollut. 159(12), 3769–3776.

van Afferden, M., Rahman, K.Z., Mosig, P., De Biase, C., Thullner, M., Oswald, S.E.,Muller, R.A., 2011. Remediation of groundwater contaminated with MTBE andbenzene: the potential of vertical-flow soil filter systems. Water Res. 45 (16),5063–5074.

Villasenor, J., Capilla, P., Rodrigo,M.A., Canizares, P., Fernandez, F.J., 2013.Operationofa horizontal subsurface flow constructed wetland–microbial fuel cell treatingwastewater underdifferent organic loading rates.WaterRes. 47 (17), 6731–6738.

Wang, X., Cai, Z., Zhou, Q., Zhang, Z., Chen, C., 2012. Bioelectrochemical stimulationof petroleum hydrocarbon degradation in saline soil using U-tube microbial fuelcells. Biotechnol. Bioeng. 109 (2), 426–433.

Wei, M., Harnisch, F., Vogt, C., Richnow, H.H., 2015. Harvesting electricity frombenzene and ammonium-contaminated groundwater using a microbial fuel cellwith an aerated cathode. RSC Adv. 5, 5321–5330.

Wu, C.H., Lai, C.Y., Lin, C.W., Kao, M.H., 2013. Generation of power by microbial fuelcell with ferricyanide in biodegradation of benzene. CLEAN – Soil Air Water 41(4), 390–395.

Yadav, A.K., Dash, P., Mohanty, A., Abbassi, R., Mishra, B.K., 2012. Performanceassessment of innovative constructed wetland-microbial fuel cell for electricityproduction and dye removal. Ecol. Eng. 47, 126–131.

You, S.J., Ren, N.Q., Zhao, Q.L., Kiely, P.D., Wang, J.Y., Yang, F.L., Fu, L., Peng, L., 2009.Improving phosphate buffer-free cathode performance of microbial fuel cellbased on biological nitrification. Biosens. Bioelectron. 24 (12), 3698–3701.

M. Wei et al. / Bioresource Technology 196 (2015) 490–499 499

Zhang, F., He, Z., 2012. Integrated organic and nitrogen removal with electricitygeneration in a tubular dual-cathode microbial fuel cell. Process Biochem. 47(12), 2146–2151.

Zhang, T., Gannon, S.M., Nevin, K.P., Franks, A.E., Lovley, D.R., 2010. Stimulating theanaerobic degradation of aromatic hydrocarbons in contaminated sediments by

providing an electrode as the electron acceptor. Environ. Microbiol. 12 (4),1011–1120.

Zhao, Y., Collum, S., Phelan, M., Goodbody, T., Doherty, L., Hu, Y., 2013. Preliminaryinvestigation of constructed wetland incorporating microbial fuel cell: batchand continuous flow trials. Chem. Eng. J. 229, 364–370.