Science of the Total Environment 533 (2015) 110–121

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Temporal evolution of bacterial communities associated with the in situwetland-based remediation of a marine shore porphyry coppertailings deposit

N. Diaby a,1, B. Dold a,2, E. Rohrbach b, C. Holliger b, P. Rossi c,⁎a University of Lausanne, Institute of Mineralogy and Geochemistry, Anthropole, Lausanne, Switzerlandb Ecole Polytechnique Fédérale de Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering, Laboratory for Environmental Biotechnology, Lausanne, Switzerlandc Ecole Polytechnique Fédérale de Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering, Central Environmental Laboratory, Lausanne, Switzerland

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

• Time-dependent sequence of majorbacterial taxa during mine tailingsbioremediation.

• Statistics showed strong links betweendiversity and ecosystem functioning.

• Populations significantly correlatedwithpH, redox, and K concentration.

• Strong dependency on constant waterflow for the stabilization of the wetland.

⁎ Corresponding author at: EPFL-IIE-GR-CEL, Station 6,E-mail address: pierre.rossi@epfl.ch (P. Rossi).

1 Laboratoire de Traitement des Eaux usées, Institut Fon2 SUMIRCO (Sustainable Mining Research & Consultanc

http://dx.doi.org/10.1016/j.scitotenv.2015.06.0760048-9697/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 10 March 2015Received in revised form 18 June 2015Accepted 19 June 2015Available online 4 July 2015

Editor: F.M. Tack

Keywords:Acid mine drainageRemediationWetlandMarine shore depositLeptospirillumAcidithiobacillus

Mine tailings are a serious threat to the environment and public health. Remediation of these residues can be car-ried out effectively by the activation of specific microbial processes. This article presents detailed informationabout temporal changes in bacterial community composition during the remediation of a section of porphyrycopper tailings deposited on the Bahía de Ite shoreline (Peru). An experimental remediation cell was floodedand transformed into a wetland in order to prevent oxidation processes, immobilizing metals. Initially, the topoxidation zone of the tailings deposit displayed a low pH (3.1) and high concentrations of metals, sulfate, andchloride, in a sandy grain size geological matrix. This habitat was dominated by sulfur- and iron-oxidizing bacte-ria, such as Leptospirillum spp.,Acidithiobacillus spp., and Sulfobacillus spp., in amicrobial communitywhich struc-ture resembled acid mine drainage environments. After wetland implementation, the cell was water-saturated,the acidity was consumed and metals dropped to a fraction of their initial respective concentrations. Bacterialcommunities analyzed by massive sequencing showed time-dependent changes both in composition and cellnumbers. The final remediation stage was characterized by the highest bacterial diversity and evenness. Asidefrom classical sulfate reducers from the phyla δ-Proteobacteria and Firmicutes, community structure comprisedtaxa derived from very diverse habitats. The community was also characterized by an elevated proportion of rarephyla and unaffiliated sequences. Numerical ecology analysis confirmed that the temporal population evolutionwas driven by pH, redox, and K. Results of this study demonstrated the usefulness of a detailed follow-up of the

CH A1 374 (Bat CH), CH-1015 Lausanne, Switzerland.

damental d'Afrique Noire (IFAN), Université Cheikh Anta Diop, BP 206 Dakar, Senegal.y EIRL), Casilla 28, San Pedro de la Paz 4130000, Chile.

111N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

remediation process, not only for the elucidation of the communities gradually switching from autotrophic,oxidizing to heterotrophic and reducing living conditions, but also for the long termmanagement of the remedi-ation wetlands.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

A total of 1150 million tons of metals have been exploited since theStone Age, generating substantial amounts of minewaste (Sheoran andSheoran, 2006). Both composition and volume of waste minerals havechanged due to the development of modern mining techniques. Nowa-days, ores are crushed and milled to reduce grain size. Records showedthat up to 80%–99% of the crushed ore may be set apart and dumped astailings waste or mixed with water and transported in suspension for afinal disposal in tailings impoundments (Sheoran and Sheoran, 2006).Acid mine drainage (AMD) forms when sulfide minerals are oxidizedon exposure to oxygen (Baker and Banfield, 2003). While the rate ofAMD formation is generally slow in massive rock assemblages, the pro-cess is accelerated as a result of grinding and mixing with chemical re-agents. However, sulfide mineral oxidation is limited as long as theground material remains water-saturated, preventing a quick access tooxygen. In unsaturated zones,microbial-enhanced oxidative dissolutionof sulfide minerals is a prime cause of water pollution (Dold andFontboté, 2001; Johnson and Hallberg, 2003). AMD formation is partic-ularly evident in the case of copper porphyry mine tailings, which typi-cally contain 0.4%–4% sulfur, mainly contained in pyrite (Dold andWeibel, 2013). In arid area such as the Atacama Desert (Peru), metalmobilization by sulfide-oxidizing bacteria result in the transportationtoward the tailings surface by evaporation. Metal precipitation occursin the form of secondary chloride and/or sulfate minerals (Diaby,2009; Dold, 2006; Dold et al., 2011; Smuda et al., 2014). These saltscan be dispersed through Aeolian transport and may contaminateadjacent areas (Dold, 2006).

Johnson (2012) showed that chemoautotrophic Bacteria and Ar-chaea are often themost numerousmicroorganisms in low pH environ-ments. The majority of these acidophilic organisms use oxygen as mainterminal electron acceptors, with Fe(III) under oxygen-limited condi-tions. Acidithiobacillus spp., Sulfobacillus spp., and Leptospirillum spp.are well known for their abilities to use reduced forms of Fe and S anduse them as sources of energy (Baker and Banfield, 2003; Johnson andHallberg, 2003; Kimura et al., 2011). Recent advances showed thatother organisms (for instance the archaean Ferroplasma spp. and thebacterial Leptospirillum Group III) were dominant in certain specificmine environments (Kuang et al., 2013; Ziegler et al., 2013). Accordingto Ňancucheo and Johnson (2012), these organisms display minimalgrowth requirements and tolerate high concentrations of dissolvedmetals in acidic solutions.

AMD formation can be avoided by limiting the contact betweenmine tailings and oxygen, for instance by submerging the tailings de-posit in a reducing environment, e.g. a constructed wetland (Johnsonand Hallberg, 2005). As a passive system, constructed wetlands areparticularly attractive due to low operating cost (Mayes et al., 2009)and are now a technology with potential success when deep reducingconditions are attained (Sheoran and Sheoran, 2006; Ňancucheo andJohnson, 2012). Constructed wetlands have been successfully appliedto somemining environments to remove contaminants fromwater, es-pecially from metal-rich effluents (Diaby and Dold, 2014). The geo-chemical processes involved in this technique are based on sulfur andsulfate reduction activities, mediated or accelerated by a heterotrophicmicrobial community fueled with organic carbon produced by theaquatic plants growing in thewetland. Sulfate reduction is a very attrac-tive method for AMD mitigation because it combines the reduction ofmetals, which is essential to their immobilization, the depletion of sul-fate anions as electron acceptor and finally the production of alkalinepotential by the production of bicarbonate ions (Moreau et al., 2013).

Sulfate-reducing bacteria have been isolated from pyrite-rich tailingsand are likely involved in sulfur cycling in these habitats (Giloteauxet al., 2013). At present, few extreme acidophilic microorganisms thatgrow via dissimilatory reduction of sulfate and/or elemental sulfur havebeen described (Johnson, 2012). Among them, most of the organismsshowed high temperature preferences, for instance the CrenarchaeotaAcidianus spp., Stygiolobus spp., Sulfurisphaera spp. or the EuryarchaeotaThermoplasma sp. and uncultured ARMAN (Archaeal Richmond MineAcidophilic Nanoorganism). Among Bacteria, recent reports men-tioned the presence of extremely acidophilic Firmicutes taxa, suchas Desulfosporosinus spp. and Desulfitobacterium spp. (Koschorrecket al., 2010; Johnson, 2012). As pH rises above 4, the diversity ofsulfate-, but also Fe(III)- and Mn(IV)-reducing acidophilic bacteria in-creases with guild members of the δ-Proteobacteria (Desulfomonilespp., Desulfobacterium spp., Desulfosporosinus spp., Desulfitobacteriumspp., and Geobacter spp.) (Koschorreck et al., 2010; Burns et al., 2011;Sánchez-Andrea et al., 2013). Other lineages were also found, such asmembers of the phylum Nitrospirae (Thermodesulfovibrio spp. andThermodesulfobacterium spp.). The presence of species traditionally be-longing to neutrophilic lineages in acidified habitats could in this casebe explained either by the presence of microhabitats or the existenceof acidotolerant strains (Sánchez-Andrea et al., 2014).

At present, little information is available on the mechanisms withwhich natural wetland systems mediate the conversion of metals andacidity on the long term (Moreau et al., 2013). Relationships betweenwetland stability, remediation capacity and the biological diversity re-main unclear (Dean et al., 2013). In this project, we present temporalchanges observed during the successful remediation of a section of amarine shoreline tailings deposit, the Bahia de Ite tailings, locatedin the Atacama Desert (Peru). Major sections of the tailings wereremediated using a constructed wetland (ca. 1000 ha) built on the oxi-dizing tailings (Diaby, 2009; Dold et al., 2011). For this study, a remedi-ation cell (ca. 30 by 30 m) was constructed on a remaining oxidizedsection of the tailings (Diaby, 2009; Diaby and Dold, 2014). Chemicalanalysis and a comprehensive survey of the microbial community pres-ent in the tailingswere conducted over a period of ninemonths, startingat the beginning of the flooding (Diaby and Dold, 2014). Goals of thisstudy were first to assess in details the time-dependent changes ob-served in the bacterial populations. Secondly, the investigation aimedto assess the significance of controlling geochemical parameters affect-ing the populations in order to evaluate possible long-term stability ofthe remediated tailings section. Careful attention was paid to the long-termpresence of facultative aerobic organisms (such as Acidithiobacillusspp.) involved in AMD formation. Elucidation of the relationships be-tween community composition and habitat evolution are anticipatedto contribute to a better understanding of the processes occurring dur-ing the rapid evolution of the remediated area. Outcomes from this de-tailed survey include the development of strategies required for theimplementation and the long term management of constructed wet-lands applied to the remediation of tailings impoundments.

2. Material and methods

2.1. Background about Bahia de Ite

Between 1960 and 1997, the Bahía de Ite (Atacama Desert, about50 km south of Ilo, Southern Peru) received about 785millionmetrictons of tailings from two porphyry copper mines, Cuajone andToquepala, operated by the Southern Peru Copper Corporation (SPCC).Grinded mine waste were sent via the Locumba River for final shore

112 N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

deposition. On average, the tailings have 4 wt.% pyrite and originatefrom a “clean ore” with very low As concentrations (around 5 mg/kgAs) associatedwith sulfideminerals. The shore tailings deposit occupiesa surface area of approximately 16 km2 and has a maximum depth of16 m at the current shoreline, located approximately 1.8 km seawardof the original shoreline. After tailings deposition ceased in 1997, SPCCinitiated a remediation campaign with the build-up of a wetlandcover over the oxidizing tailings that applied alkaline water from theLocumba River (Diaby, 2009; Dold et al., 2011).

2.2. Constructed wetland

A remediation cell (ca. 30 × 30 m) was constructed in September2005 close to the Northern section of the Bahia de Ite tailings, ca. 50 maway from the shoreline, at sea level (SD Photographies 1–3). This tail-ings section displayed a heterogeneous stratigraphy consisting of main-ly sand horizons (Diaby, 2009; Diaby and Dold, 2014). The surroundingdikes (ca 70 cm high) were constructed manually using adjacent tail-ings material. A piezometer nest was installed in the center of the cell.The cell was flooded using the water channeled from an adjacent sec-tion of the tailings that were already remediated (Table SD-1: chemicalcomposition of the water at start). As the deposited tailings materialwas relatively coarse, the water used for the remediation infiltratedquickly. The cell was fully water-saturated after 6 days of continuousflooding. Once the remediation cell was saturated, the depth ofthe water cover was maintained at about 50 cm. At D30, local flora(Typha sp.), was planted on the remediation cell, with small amountsof straw and manure deposited as organic fertilizers.

2.3. Sampling

Water samples were collected from a piezometer nest located at thecenter of the constructed cell at regular frequency using a peristalticpump, screened at a depth of 30 to 40 cmbelow tailing surface. Samplesdestined for molecular analysis were composed each of 1 l of water,mixed with fine sandy to silty sediment particles. Samples were storedin sterile bottles and kept on ice in the dark for transportation to the lab-oratory. The samples were filtered aseptically through a sterile 0.2 μmregenerated cellulose filter (Millipore) in the laboratory. Filters wereimmediately frozen and kept at −20 °C until processing. Samplesused for chemical analysis were filtered (0.2 μm, regenerated cellulose)and stored at 4 °C in the dark prior to analysis. Samples used for cationanalyses were acidified with suprapure HNO3. Geochemical andminer-alogical techniques and results are presented in details elsewhere (Doldet al., 2011; Diaby and Dold, 2014).

2.4. DNA extraction and amplicon library preparation

Total DNA was extracted from filters using the PowerSoil DNA ex-traction kit (MoBio, Carlsbad, USA). DNA samples were quantified(QuBit 2.0 and QuBit dsDNA HS Assay Kit, Life Technologies) and werestored at −25 °C prior to analysis. PCR amplification of the 16S rRNAhypervariable regions V1–V3 was carried out using bacterial HPLC-purified primers 28f and 519r (5′-GAGTTTGATCNTGGCTCAG-3′ and5′-GTNTTACNGCGGCKGCTG-3′ respectively), with 0.1 ng/μl of templateDNA (final concentration). Ca. 520 bp ampliconswere generated in 50 μlreaction volumes containing 5 μl of 10× PCR Buffer, 5 μl of 25 mMMgCl2, 9 μl of Enhancer P, 3.6 μl of dNTPs (2.5 μM each), 1 μl of eachprimer (10 μM) and 0.5 μl of PeqGold DNA polymerase (all PeqLab,Germany). PCR amplification conditions were as follows: 94 °C for5 min, followed by 25 cycles of 94 °C for 30 s, 56 °C for 30 s, 72 °C for60 s, and a final elongation step of 72 °C for 5 min. Amplicons purifica-tion was carried out with magnetic beads (AxyPrep Mag PCR Clean-Up, Axygen, USA). Amplicons were reduced in size using the enzymesprovided in the Ion Xpress Plus Fragment Library Kit and were fusedwith both adaptors A and the barcoded Pi (Ion Xpress Barcode Adapters

1–16 Kit) according to themanufacturer instructions (all Life Technolo-gies). Size selection (max370 bp)was carried out on agarose gels (E-GelSystem, Life Technologies). Quantification of the fragments was carriedout using a BioAnalyser 2100 and the High Sensitivity DNA chips(Agilent technologies, USA).

2.5. Semiconductor sequencing

Emulsion PCR was carried out applying the Ion XPress Template Kit(Life Technologies) as described in the appropriate User Guide providedby the manufacturer. Sequencing of the amplicons was carried out onthe Ion Torrent Personal Genome Machine (PGM) using the IonSequencing 300 kit (Life Technologies) equipped with a 316 chip fol-lowing the corresponding protocol. Pooled barcoded samples wereloaded on the same chip.

2.6. Sequence analysis

Primary base callingwas performedusing Torrent Suite v3.0 software(Life Technologies). Sequencing readswere downloaded as .sff files fromthe Torrent Server and were processed on a Linux Ubuntu platform(BioLinux 7, (Field et al., 2006)) running on a local Dell PrecisionT3600 2 GHz bench top computer, equipped with a 12 cores processorarray and 32 GB of RAM. Reads were consecutively checked for differentquality criteria inMothur (Schloss et al., 2009). Stringent selection of thereadswas carried out according to i) the size of the amplicons (N150 andb300 bp) ii) the absence of ambiguous called base (N), iii) an averagePhred quality score N25 and iv) the absence of homopolymers longerthan 8 bp. Reads were aligned with the Greengenes database (DeSantiset al., 2006) and de-noised with the Single Linkage Preclustering (SLP)method (Huse et al., 2010). In this process, reads having a pairwise dis-tance smaller than 2%were clustered andmergedwhile retaining clustersizes as counts (Jünemann et al., 2012). Reads were further screened forartificial chimeric formations using the UCHIME algorithm implementedin Mothur (Edgar et al., 2011) and the SILVA database (Quast et al.,2013). Taxonomy was assigned to the lowest possible rank to the de-rived high quality reads using the latest version of the Greengenes data-base. Unknown reads and reads affiliatedwith theDomainArchaeawereremoved. Hierarchical clustering was carried out using Esprit-Tree (Caiand Sun, 2011). The outputwas used to computeOperational TaxonomicUnits (OTUs), rarefaction curves, as well as Abundance-based CoverageEstimators (ACE) and CHAO1 estimator, which assessed the putativetotal richness by adding a correction factor to the observed number ofspecies (Hughes et al., 2001). The freeware R (R Development CoreTeam, 2009) was used subsequently for numerical ecology analysisand inference statistics, for the computation of Fishers's α index (asemi-parametric index independent from the size of the sampling)and Pielou's evenness. Multifactorial analysis (MFA) was conductedon environmental and microbial data sets. MFA is a symmetrical analy-sis, in which both data sets play the same role, aimed at revealing an ex-ploratory point of view. This analysis exposes correlative structureswithout any reference to causal relationship (Borcard et al., 2011).

2.7. Data availability

Raw sequences were deposited in the European Nucleotide Archive(http://www.ebi.ac.uk/ena) and are available under the sample groupID ERG004003.

3. Results and discussion

3.1. Strong temporal evolution of the chemical and physical parametersfollowed the flooding of the constructed remediation cell

Physico-chemical data have been initially presented in Diaby (2009)and published more recently in Diaby and Dold (2014). These data are

113N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

summarized for the convenience of the reader in Table SD-1 and Fig. SD-1, as they were used here for correlation analysis with microbiologicaldata. In oxidizing tailings environments, geochemical patterns are typi-cally characterized by strong vertical gradients (Dold, 2014). IncreasingpH and decreasing redox potential with depth are the consequences ofthe chemical oxidation of sulfide minerals (e.g. pyrite) in the presenceof oxygen, a reaction catalyzed and accelerated by microbial activity(Johnson and Hallberg, 2003). At the start of the remediation process,the surface of non-remediated tailings displayed an accumulation ofwater soluble efflorescent salts, developed through upward capillarytransport in a hyper-arid climate (Diaby, 2009; Diaby and Dold, 2014).At a depth of about 40 cm below the top of the tailings, the oxidationzone was characterized by low pH (3.1) and by a redox potential of358 mV (Fig. 1). Pore water samples taken from this zone were rich inmetals, with the following concentrations: 12.9 mg/L Zn, 80.1 mg/LMn, 104 mg/L Al, 176 mg/L Fe, 329 mg/L Cu (Diaby and Dold, 2014).Cl, Na and SO4 concentrations were respectively 2125 mg/l, 1748 mg/land 4167 mg/L, suggesting that the tailings were mainly saturatedwith fresh water with very little seawater infiltration. Below the oxida-tion zone of the non-remediated tailings, primary tailings deposits,which were protected from any oxygen intrusion, displayed higher pHvalues (5.0 to 6.7) and lower redox potentials (128 to 225 mV).

Once remediation (flooding) started, pH decreased initially in theoxidation zone to 2.5; it rebounded to 5.9 after 3 weeks and 6.5 after 5months of remediation. Na and Cl concentrations decreased shortlyafter the remediation started, most likely due to the increased hydraulicgradient and the subsequent dilution, which pushed the seawater–freshwater interface toward the sea (Diaby, 2009; Dold et al., 2011).Eh values decreased from 358 mV to 300 mV after 3 weeks, reaching135 mV after 5 months. These more reducing conditions and greateravailability of organic matter from both plant rhizosphere and theadded manure contributed to the development of favorable growthconditions for metal- and sulfate-reducing bacteria (MRB, resp. SRB).Changes in pH values as well as the formation of more reducing condi-tions triggered a decrease of metal concentration with time. As a conse-quence, Fe, Mn, Al, Cu, and Zn concentrations were close or belowdetection limit after 3 weeks of continuous flooding (Fig. SD-1). Thecontinuous flooding of the constructed cell was perturbed at a certainpoint by a lack of incoming water during the dry season. At D77, lowsurface water input (dryness in summer) triggered a lower hydraulicgradient in the wetland and subsequently resulted in a seawater intru-sion, visible by the increase of Na, Cl, and sulfate during this period(Diaby and Dold, 2014).

Standardized Sampling Effort0.0 0.2 0.4 0.6 0.8 1.0

OT

Us

- 5%

clu

ster

ing

leve

l

0

4000

8000

12000

16000

20000D0D7D13D19D47D77D104D169D225

Fig. 1. Rarefaction analysis carried out on the bacterial communities sampled from D0(flooding of the constructed remediation cell) to D225 using ESPRIT-Tree (5% clusteringlevel).

3.2. Bacterial populations showed rapid increase in diversity and evennessas a response to drastically changing environmental conditions

The numbers of OTUs obtained at 5% clustering level varied substan-tially among samples. OTUs varied from 4208 for sample D7 to a maxi-mum of 17,889 for sample D285 (Table 1). The specificity of sample D7is confirmed with the lowest values computed for both ACE and CHAO1estimators, as well as extreme values for both Fishers's α index andcommunity evenness (computed with Pielou's estimator). Populationssampled at D7 were strongly impacted by the incoming water, whichresulted in a community structure showing a low number of highlydominating taxa.

From sample D7 on, the analysis of the community showed a long-lasting evolution of the structures over time. The major changes werereflected by the increase in community diversity with time. Rise of thediversity over time is expressed using rarefaction curves (Fig. 1). Con-comitantly, the analysis showed a steady diminution of the sequencesaffiliated with dominant species normally associated with AMD forma-tion. Leptospirillum spp. and Sulfobacillus spp. respective contributionsdecreased substantially and composed less than 0.3% of the final com-munity measured by the end of the experience.

The AMD-impacted samples D0 and D7 initially showed the lowestdiversity (with a respective 4620 and 4208 OTUs), the lowest evennessand the lowest percentages of unaffiliated sequences (Table 1). Floodingthe remediation cells induced a gradual increase of the diversity withtime (Fig. 1), with high ACE and CHAO1 values reached at D77 firstand then at D225 with 17,889 OTUs (Table 1). D225 showed the secondhighest evenness and the highest percentages of unaffiliated sequences,with 40.5% of all sequences (Table 1). A high proportion of unaffiliatedsequences is not uncommon in complexmicrobial habitats, and the rel-ative proportion reaching up to 38% were revealed during the remedia-tion of contaminated marine sediments (Dell'Anno et al., 2012). Apossible explanation for the relatively high proportion of unaffiliatedsequence may come from the juxtaposition of different habitats. Theremediation cell was located near the seashore, and was connected toformerly remediated sections of the tailings. The water used for the re-mediation was already impacted from its passage from high mountainsto the ocean, via desert areas and agricultural lands, while receivingcontributions from human activities, such as agriculture and industry.One consequence of the seeding of allochthonous bacterial taxa in theremediation cell was the increase of diversity and the concomitantrise of unaffiliated bacterial sequences with time (Fig. 2).

Clustering of the bacterial communities, present in all nine samples,showed the formation of threewell-defined groups, following themod-ification of the main environmental factors with time (Fig. 3). The firstclusterwas formedwith sampleD0 andD7, at the initiation of the reme-diation procedure, confirming preceding results. The second clustergrouped samples D13 to D47, indicating an intermediate or transitionalphase of the remediation. Finally the third cluster was formed by thesamples taken from D77 to D225 and formed a sequence duringwhich the bacteria-mediated remediation process deployed its full ef-fectiveness. Interestingly, identical clustering patterns, showing the for-mation of identical groups, were found at all phylogenetic levels, fromphylum to species (Fig. SD-2). The coherence shown by the clusteringtechnique at all levels displayed evidence of biological processes thatoccurred with bacterial actors linked to deep phylogenetic origins.This result indicated that taxa involved in the remediation processfollowed a strong niche differentiation with time between deeplydiverging groups (Philippot et al., 2010).

3.3. Initial stage of the remediation process, from D0 to D7

The bacterial community present in the initial oxidizing tailings (D0)was composed by a limited number of phyla (20). Among these phyla,Nitrospirae (41.0%), Proteobacteria (22.6%), Firmicutes (13.8%),Actinobacteria (6.7%), Bacteroidetes (5.6%), and Chloroflexi (2.8%)

Table 1Estimators computed with high quality reads at 5% clustering level using Esprit-Tree and R.

Sample days Readnumbers

Phylumrichness

OTUs ACE CHAO1 Pielou'sevenness

Fisher's α Unaffiliatedsequences (%)

D0 73,122 20 4620 6371 6670 0.55 50.55 6.5D7 64,445 20 4208 6011 6155 0.63 36.31 12D13 102,667 26 7329 11,220 12,000 0.46 45.82 23.6D19 79,651 22 5928 9227 9733 0.52 37.69 15.5D47 100,854 31 7808 12,910 13,903 0.50 37.57 15.6D77 142,910 37 15,552 27,622 30,241 0.50 63.39 30.5D104 157,977 33 10,707 15,886 16,588 0.52 47.56 28.5D169 117,196 35 11,175 17,904 19,078 0.48 44.25 41.7D225 120,064 42 17,889 33,958 36,815 0.48 61.91 40.5Mean 106,543 30 9468 15,679 16,798 0.52 47.23 24

114 N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

showed high respective contributions only, and composed ca. 92.4% ofall sequences. With the addition of the unaffiliated sequences this pro-portion reached 98.9% of all sequences, leaving ca. 1% to minor phyla(Fig. 4 and Table SD-5).

The initial dominance of Acidithiobacillus spp., Sulfobacillus spp., andLeptospirillum spp. (reaching up to 53% of the total sequences found inD0,) was demonstrated already in other habitats and is characteristicfor bacterial communities linked with AMD generation, as they par-ticipate to the oxidation of sulfide minerals (Chen et al., 2013).AMD-related genus Acidiphilium spp. and Acidisphaera spp. (phylumProteobacteria) composed ca. 1.5% and 0.1% respectively of all se-quences. Among the phylum Firmicutes, a quick rise in relative propor-tions of Alicyclobacillus spp. (Order Bacillales) occurred immediatelyafter the start of the flooding (Fig. 8). This genus, composed of strictlyaerobic and acidophilic species, was found to be an essential elementlinked toAMD formation in former studies (Korehi et al., 2014). Surpris-ingly, other taxa found classically in AMD-forming communities, suchas Thiomonas spp., Acidocella spp. (phylum Proteobacteria), andAcidimicrobium spp. (phylum Actinobacteria) were absent in all sam-ples of this study (Volant et al., 2014).

pH values at D7were lower than thosemeasured at D0 (2.4 vs 3.1 re-spectively), whilemetal concentrations dropped as a consequence of thedilution and dispersion of the cations (Table SD-1). Flooding the experi-mental cell with neutral pH water impacted the members of the AMD-forming bacterial guild. Flooding impacted Leptospirillum ferriphilum,Acidithiobacillus albertensis, and Sulfobacillus thermosulfidooxidans,which relative contributions reached almost zero at D7 (Fig. 5). Con-versely, other strains of iron- and sulfur-oxidizers were favored duringthe first few days of the flooding, apparently taking advantage of lowerpH values, namely Leptospirillum ferrodiazotrophus, Alicyclobacillusferrooxydans, unclass. Acidithiobacillus spp., and unclass Sulfobacillus spp.

At D0, 14.8% of the initial community was composed of bacteriaaffiliated with α- and β-Proteobacteria sub-phyla. These taxa

Samples

D0 D7 D13 D19 D47 D77 D104D169D225

Rel

ativ

e C

ontr

ibut

ions

[%]

0

10

20

30

40

50

CH

AO

1 Estim

ator

0

10000

20000

30000

40000

50000

60000

70000

80000

Fig. 2. Evolution of the relative contributions of unaffiliated bacterial sequences with time.Black dots: CHAO1 estimators.

were dominated by members of the orders Sphingomonadales,Burkholderiales, and Rhodocyclales, usually associated with soil, plantroot and animal habitats, such as Shingomonas spp. (4.3%) and Zoogleaspp. (3.3%) (Fig. 6). Both taxa were traditionally also associated withwetland ecosystems, making up to 10% of the community members(Hartman et al., 2008). In this study however, both sub-phyla accompa-nied negatively the remediation process and composed only a smallfraction of the community in D225, with respectively ca. 0.6% and 1.2%of all sequences (Fig. 6). Despite numerous findings implying the pres-ence of Gallionella spp. in AMD-impacted habitats, members of thisiron-oxidizing clade were found in D77 only, contributing with lessthan 0.1% to the global community.

A limited number of rare taxawere detected in the initial D0 sample,composing ca. 1% of the sequences (Fig. 5). These phyla were identifiedpossibly as a result of the extensive analysis carried out by deep se-quencing, which provided a much higher genetic diversity than previ-ous estimates based on clone libraries. Whereas low-abundance taxamay account for stability and function of communities (Sogin et al.,2006), a recent study showed that members of this rare biosphereplayed a potentially critical role in AMD formation by promotingcomplementary functions such as nitrogen fixation (Hua et al., 2014).Conversely, photosynthetic microorganisms (Cyanobacteria spp. for in-stance), by consuming a weak base (bicarbonate) and producing astrong base (hydroxyl ions), were found to generate net alkalinity, rein-forcing remediation processes (Johnson and Hallberg, 2005). It isunclear however, whether all members from rare taxa contributed toeither AMD formation or the remediation process, or if they werereflecting allochthonous inputs only. For instance, the presence ofspecific sequences linked with bacterial-plant relationship (such assequences affiliated with Novosphingobium spp. or Burkholderia spp.)could be interpreted as an indication of the groundwater flow fromthe already remediated sections of the upstream wetland (Diaby andDold, 2014). Other sequences, such as those affiliated with the marineProchlorococcus spp., denoted possible impact of the marine environ-ment (tide, waves and strong winds). In this sense, the diversity foundin the initial sample did not reflect fully the composition of the autoch-thonous community and was overestimated by additional inputs fromadjacent habitats.

Finally, small but significant numbers of MRB and SRB were de-tected in the initial sample D0, an indication of the complexity and

D16

9

D22

5

D77

D10

4

D47

D13

D19 D

0

D70.

20.

61.

0

Hei

ght

Fig. 3. Minimum variance clustering on Hellinger-transformed bacterial data sets at thephylum level, presented with the Ward technique.

Actinobacteria6.7%

Bacteroidetes5.6%

Chloroflexi2.8%

Firmicutes13.8%

Nitrospirae41.0%

Proteobacteria22.6%

Unclass. 5.6%

ABY1OD1

Acidobacteria

Cyanobacteria

Elusimicrobia

Fusobacteria

GN02

OP3

Planctomycetes

TM6TM7

Tenericutes

Thermi

Verrucomicrobia

Minor phyla : 1.03 %

Fig. 4. Phylogenetic assignment of the sequences obtained for sample D0, before the initiation of the flooding of the constructed remediation cell.

115N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

structuring of the initial habitat (Fig. 7). Despite conditions appar-ently unfavorable for such activities, taxa affiliated to the phyla γ-Proteobacteria (Geobacter spp., Desulfovibrio spp.), ε-Proteobacteria(Sulfurospirillum spp.) and Firmicutes (Desulfosporosinus spp.,Dehalobacter spp. and Desulfitobacterium spp.) contributed fromless than 0.1% to 0.2% to the total sequences. This finding are consistentwith results obtained by other researchers, whodemonstrated the pres-ence of SRB in acidic water discharges from mining activities (Alazardet al., 2010).

10

20

30

40L. ferriphilumL. ferrodiazotrophusUnclass. Leptospirillum sp.

2

4

6

8

10A. albertensisA. caldusA. ferrooxidansUnclass. Acidithiobacillus sp.

SamplesD0 D7 D13 D19 D47 D77D104 D169 D225

Spe

cies

con

trib

utio

ns [%

]

0

1

2

3

4

5

6S. acidophilus S. benefaciensS. thermosulfidooxidansUnclass. Sulfobacillus sp.

Fig. 5. Temporal changes observed for Leptospirillum spp. (top), Acidithiobacillus spp.(middle), and Sulfobacillus spp. (bottom) relative contributions in the communities pres-ent in the constructed remediation cell.

3.4. Second phase of the remediation process, from D13 to D47

According to Diaby and Dold (2014), the observed rise of pH valuesduring this second phase could be attributed to the flushing of dissolvedelements and the reduction of sulfate. The generation of hydrogen sul-fide triggered the precipitation of metals from solution in insolublecomplexes. Dissimilatory reduction of sulfate into sulfide generated al-kalinity by transforming a strong acid into hydrogen sulfide, which con-tributed potentially to the rise of pH values (Johnson and Hallberg,2009). Removal of metals from the solution by the activity of SRBguild members in AMD was confirmed in other studies (Johnson andHallberg, 2005; Dold et al., 2011; Dean et al., 2013; Moreau et al.,2013). In this study, dominant SRB guild members were associatedwith phyla δ-Proteobacteria and Firmicutes (Fig. 7). Both bacterialphylamembers found favorable growth conditions, such as low concen-trations of oxygen, DOC (up to 17 mg/L), electrons acceptors(1829 mg/L sulfate) and Fe(III) hydroxide (mainly as ferrihydrite, goe-thite and hematite) (Diaby, 2009).

At D19, a total of 27.6% of sequences were affiliatedwith the phylumProteobacteria. This phylum showed very high proportions of unaffiliat-ed sequences, culminating with 7.9% at D169. Interestingly, sequencesaffiliated with the order Desulfovibrionales (Desulfovibrio spp.) weredetected in sample D7 already, having a relative contribution of 0.8%.The early settlement of these guild members could be possibly linked

RhizobialesRhodobacteralesRhodospirillalesSphingomonadalesUnclass - ProteobacteriaBurkholderialesHydrogenophilalesRhodocyclalesUnclass - Proteobacteria

SamplesD0 D7 D12 D18 D47 D77 D104 D169 D285

Rel

ativ

e co

ntrib

utio

ns [%

]

0

3

6

9

12

15

Fig. 6. Temporal changes of the relative contributions of selected orders in the sub-phylaα- and β-Proteobacteria during the remediation process.

2

4

6

8

10

12DesulfobacteralesDesulfovibrionalesDesulfuromonadalesSyntrophobacteralesUnclass. -Proteobacteria

SamplesD0 D7 D13 D19 D47 D77D104D169D225

Tax

a co

ntrib

utio

ns [%

]

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5Desulfitobacterium spp.Desulfosporosinus spp.Thermincola spp. unclass. Peptococcaceae

Fig. 7. Temporal changes observed in the relative contributions of selected orders of the δ-Proteobacteria (above) and in selected genus of the Firmicutes (below). AmongFirmicutes, SRB guild members showing relative contributions above 0.1% are displayedonly.

5

10

15

20

25

30Bacteroidales Flavobacteriales Sphingobacteriales Unclass. Bacteroidetes

4

8

12

16BacillalesLactobacillalesClostridialesUnclass. Firmicutes

Tax

a co

ntrib

utio

ns [%

]

SamplesD0 D7 D13 D19 D47 D77 D104 D169 D225

0

1

2

3

4

5

6AcidimicrobialesActinomycetalesWCHB1-81 Unclass. Actinobacteria

Fig. 8. Temporal changes observed in the relative contributions of selected orders amongthe Phyla Bacteroidetes (top), Firmicutes (middle), and Actinobacteria (bottom) in theconstructed remediation cell.

116 N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

with the inherent capacity to tolerate the presence of oxygen andwith the oligotrophic status of the habitat. Members of the orderDesulfuromonadales, represented mainly by unclassified sequences,contributed to 7.2% at D19, and peaked with 10.1% of all sequences atD77 (Fig. 7). Among this Order, Geobacter spp. relative contributionculminated with 2.6% in D13, whereas the maximal contribution forPelobacter spp. was attained later with 3.3% at D77. Both contributionswere in accordance with traditional dominance found in other AMD-contaminated environments (Petrie et al., 2003). At D225 however,these two genera contributed respectively with ca 0.1% and 0.9% of allsequences only. The decline of these two genera could possibly belinked with the large dominance of unclassified Desulfuromonadales,outcompeting known organisms with their relatively high contribu-tions (ca. 6.5% at D77 and 1.6% at D225). Within the same order,Desulfuromonas spp. and Desulfuromusa spp. contributed marginallyto the communities throughout the study.

Members of the orders Desulfobacterales (Desulfobulbus spp.Desulfocapsa spp., Desulforhopalus spp. and Desulfotalea spp.) weregradually involved from D47 onwards, making them the last SRB guildmembers to settle in the remediation system. The highest contributionof this order was from unclassified sequences making up to 2.54% ofall sequences at D225. Relative contributions of sequences affiliatedwith the Syntrophobacterales were increasing steadily, starting withdetectable sequences at D13, and reaching 1.3% at D225. Unaffiliatedsequences and Desulfobacter spp. composed the vast majority ofsequences affiliated with this order, whereas Desulfobacterium spp.,Desulfobacula spp., and Desulfococcus spp. were detected in marginalproportions only (b0.1%).

Among Firmicutes, only sequences affiliated with Desulfosporosinusspp. and to a lesser extend Desulfitobacterium spp. (order Clostridiales)contributed significantly to the communities (Fig. 7). Desulfurosinusmeridiei was detected in sample D13 and this species contributed to amaximum 2.4% of all sequences in sample D19. Desulfosporosinus spp.was already isolated from acidophilic habitats impacted by AMD at pHvalues ranging from 3.6 to 5.5 (Alazard et al., 2010) and in othermetal-contaminated environments, in which a possible activity in pres-ence of traces of oxygen could have been possible (Karnachuk et al.,2009). These strictly anaerobic bacteria are spore-forming, providing a

strong ecological advantage in constantly perturbed environments.Sequences affiliated with Thermincola spp. were the second most abun-dant among the Clostridiales, detected from D47 on and peaking atD169 with 0.8%. This organism is traditionally considered as a metal-reducing bacterium (MRB) and is moderately thermophilic, spore-forming, and able to reduce amorphous Fe(III)-hydroxides (Zavarzinaet al., 2007). Other Clostridiales (Dehalobacter spp.), aswell asmembersof the Desulfitobacterales were detected from D19 on, but contributedonly slightly (b0.1%) to the communities throughout the whole re-mediation process. A large contribution was provided however byunclassified Clostridiales sequences, peaking at D19 with a relativecontribution of 3.3%.

From D13 on, the environmental conditions prevailing in theremediated tailings induced a shift from acidophilic tomore neutrophil-ic organisms. This shift was also visible in the strong development ofheterotrophic organisms from specific phyla, such as Bacteroidetes,Firmicutes (others than SRB members), and Actinobacteria (Fig. 8).These three phyla contributed with respectively 35.7%, 8.3% and 1.1%of all sequences found at D47, confirming former studies. Korehi et al.(2014) found that both Bacteroidetes and Firmicutes were dominantin moderate acid conditions. In the present study, the high relative con-tributions of these heterotrophic phyla coincidedwith the developmentof the planted Typha sp. as well as the deposition of manure and strawwithin the remediation cell, providing new sources of organic carbon(Table SD-2). Members of the Bacteroidetes phylum are considered tobe r strategists, growing rapidly under conditions of high resourcesavailability (Philippot et al., 2010). They were considered as primarymetabolizers of labile carbon inputs commonly associated with soilsrich in organic carbon (Fierer et al., 2007). In this study, Flavobacteriales,Sphingobacteriales, and Bacteroidales composed the three major

SamplesD0 D7 D13 D19 D47 D77 D104 D169 D225

Rel

ativ

e co

ntrib

utio

ns [%

]

0

1

2

3

4

5

ABY1OD1 ArmatimonadetesBRC1 ChlorobiElusimicrobiaGN02 GemmatimonadetesLentisphaeraeNKB19 SynergistetesTM6 Verrucomicrobia

Fig. 9. Temporal changes of the relative contributions of rare and candidate phyla duringthe course of the remediation process.

117N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

Families of this phylum, with the highest proportion being representedby unclassified sequences (27.7% at D19, Fig. 8).

An important phylum that developed during this phase of the reme-diation is the Acidobacteria, with a maximum contribution of ca. 6.2% atD47. Bacterial members associated with this phylum are K strategistsand have been found in sedimentary ecosystems, soils, freshwater aswell as marine sediments and deep terrestrial sub-surfaces (Coateset al., 1999). Among this phylum, Geothrix spp. (Order Holophagales)contributed with a maximum of ca. 4.4% of all sequences at D47, declin-ing to less than 0.1% at D225. Geothrix spp. are MRB, coupling the oxida-tion of organic compounds with the reduction of ferric iron to supportgrowth (Lovley, 2013). Members of the Holophagales are favored atacidic pH values (Korehi et al., 2014). In this study, they wereoutcompeted later on when pH values stabilize near neutrality. Thepresence of Geothrix spp. composed a typical marker of this secondphase of the remediation process.

3.5. Final phase of the remediation process, from D77 to D225

Three weeks after the initiation of the remediation process, the re-versing of the reactions leading to the generation of AMD had becomeobvious (Fig. 1). Changes in environmental variables started to slowdown and to stabilize, with pH values around 6.5 and with metal con-centrations below detection limits. Eh values decreased from 292 mV(D77) and reached 176 mV at D225. Fe was the only element presentin significant amounts, showing concentrations up to 13.7 mg/L atD225 (Table SD-2). As mentioned previously, the increase in the sulfateconcentrations during the dry season at D77 was a consequence of theapparent rise of the freshwater/seawater interface, induced by the de-crease of ca. 0.5 m of the wetland water level. After D169, the decreaseof sulfatewas interpreted as the consequence of sulfate reductionmedi-ated by bacterial activities, with the concomitant rise of sulfide concen-tration from 1.6% to 3.2% (Diaby and Dold, 2014).

This third period of the remediation process was characterized by astrong decline of themetal- and sulfur-oxidizers. At D77, less than 1% ofthe total sequences were affiliated with members of the AMD-formingbacterial guild. Leptospirillum spp. and Acidithiobacillus spp. showed arelative contribution below 0.1% at D225 (Fig. 5). An exception occurredwith unclassified sequences affiliated with Sulfobacillus spp. andAlicyclobacillus spp. (Firmicutes), rebounding at D77 and D104,confirming the occurrence of an environmental perturbation in the re-mediation process (sea water intrusion). The ability of these taxa toform spores and therefore to resume active growth when more favor-able conditions were restored, conferred the capacity to be among thefirst bacterial cells linked with the reversal of the remediation process.Furthermore, the versatility of their metabolic functions may provideother reasons for which these taxa were still present at the end of thisstudy. Although these contributions may look insignificant in terms ofrelative contributions, the long term preservation of members of thisguild has a deep impact on the long term management of the wetlandsystem. A quick reversal of the remediation process, induced by a periodof drought for instance,would still be possible. The system remains frag-ile and the constant provision of incoming river water is essential.Among SRB andMRB, this stage of the remediation started with the de-cline of the sequences affiliated with the Firmicutes (Desulfosporosinusspp.), whichwere traditionally linkedwithmore acidic and oligotrophicenvironmental conditions (Fig. 7). From D104 on, main SRB activitieswere driven apparently by δ-Proteobacteria members. The large domi-nance of sequences affiliated with the order Desulfuromanadales waschallenged by the constant increase in sequences affiliated withDesulfobacterales and Syntrophobacterales (2.5% and 1.3% respectivelyat D225).

Asmentioned before, analysis of the sequences showed a gradual in-crease of the diversity with time during the remediation process(Table 1). In this study, 18 rare phyla contributed to less than 0.1% ofall sequences, while 14 contributed between 0.1 and 1% of all sequences

(Table SD-5). This fraction of the communities, that initially representeda minority of the biomass, branching often deeply within the bacterialtree, has been described as the rare biosphere (Sogin et al., 2006).Even though the relative contributions of these phyla were low, the af-filiation of specific sequences could provide useful information aboutthe progress and the functioning of the remediation process (Fig. 10).For instance, Chlorobi (principally unclassified members of the OrderZB1) reached amaximum relative contribution of ca 1% in D77. The can-didate lineage ZB1was defined by Elshahed et al. (2003) from sequencesisolated from a microbial mat in a sulfide–sulfur rich artesian spring.The vertical migration of these non-motile photoautrophic green sulfurcells from the surface to ca 40 cmbelow the tailings level could be an in-dication of an anaerobic phototrophic activity within the remediationcell. Another rare phylum was composed by the candidate phylumTM6, contributing with 0.9% of all sequences at D225. A large numberof TM6-related 16S rRNA gene sequences have been identified fromgeographically varied sampling sites, which suggests that this phylumhas a widespread distribution, including AMD biofilms, although typi-cally found at low relative abundance (McLean et al., 2013) (Fig. 9).

3.6. Correlations between environmental factors and bacterial communities

Multifactorial analysis (MFA) was carried out on both environmen-tal and massive sequencing data sets (Fig. 10). Results showed thatthe two first axes represented more than 68% of the total variance.Axis 1 expressed the largest amount of the variance (47.2%), with theclose association of vectors corresponding to the variables K, pH, sulfate,metals, and redox, showing high RV coefficients (Table 2). Interestingly,RV coefficients showed that almost identical variableswere significantlycorrelated at all phylogenetic levels, to the exception of total Fe thatwasfound significant at the phylum level only. These findings confirmed thecoherence shown by the clustering technique at all levels (Fig. SD-2)that indicated a strong link between very specific biological activitiesand bacterial actors branching at deep phylogenetic levels. In otherworlds, the taxa involved in the remediation process followed a strongniche differentiation with time, involving the implication of deeplyrooted bacterial guilds (Philippot et al., 2010).

On Fig. 11 (right panel), the opposite directions of vectors K and pHon one side and the redox vector on the other fully confirmed that eco-logical conditions prevailing during the remediation of the artificialwetland were dominated by variables impacting strongly the physio-logical functioning of the bacterial communities. pH values wereshown already to be one of the most important factor impacting andshaping bacterial community structures in AMD-impacted habitats

-2 -1 0 1 2 3 4

-10

12

Dim 1 (47.2 % of variance)

Dim

2 (

20.8

% o

f va

rian

ce)

D0

D7D12

D19

D47

D77D104

D169

D225

speenv

-1.0 -0.5 0.0 0.5 1.0

-1.0

-0.5

0.0

0.5

1.0

Dim 1 (47.2% of variance)

Al

redox

CuZnMn

FeII

Fetot

K

pH

SO4

Dim

2 (

20.8

% o

f va

rian

ce)

Fig. 10. Left: multi factorial analysis (MFA) site scores using bacterial phylogeny at the species level linked to environmental datasets. Open circles: individual scores for both bacterialspecies and environmental data sets; black circles: sample score centroid. Right: correlations between environmental variables and theMFA site scores on Axes 1 and 2. Variables showingstatistically significant RV coefficients (p b 0.05) are displayed only.

118 N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

(García-Moyano et al., 2012; Chen et al., 2013; Liu et al., 2014). From astrict ecological point of view, bacterial taxa branching at high phyloge-netic levels are strictly infeoded with limited ranges of pH preferences,developing specific metabolic functions, and making major phylamore predictable. Increasing concentrations of K with timewere signif-icantly correlatedwith the rise of pH values and showed a RV coefficientof −0.873 (p = 0.002**) at the species level (Table 2). Low concentra-tions of K (21 mg/L) were present at the beginning of the remediationprocess already (Table SD-2). These concentrations doubled andreached 54.6mg/L at D225, reaching the concentrations found in the in-comingfloodingwater. Recent advances showed no limitation impact oflow concentrations of K on the growth and diversity of microbial com-munities (Moro et al., 2014). On the other hand, a clear impact on litterdecomposition was found in a globally micronutrient-limited Amazo-nian rain forest soil (Kaspari et al., 2008). In this study, and as themicro-nutrient sourcesweremanifold and abundant, the relationship betweenK and the communities was interpreted as a strong indication of a con-tinuum in the evolution of the populations accompanying the remedia-tion process. Sulfate was the only variable significantly correlated withAxis 2, showing a RV coefficient of 0.829 (p = 0.006**). This variablewas shown already to be responsible for the shaping of AMD-relatedbacterial communities (Volant et al., 2014). Interestingly, this vectorwas orthogonal with the vector associated with Axis 1. This can beinterpreted as a variation of sulfate concentrations that were totallyindependent from the changes of pH and redox values.

Table 2RV Coefficient computed with MFA using individual variable at three phylogenetic levels. Sign

Phylum level 0.05 Order le

RV RV

Variables Coefficients p-Values Coefficie

pH 0.846 0.004** 0.882K 0.880 0.002** 0.888FeII −0.803 0.009** −0.765Mn −0.801 0.009** −0.768Zn −0.797 0.010** −0.764Cu −0.809 0.008** −0.778Al −0.834 0.005** −0.806Redox −0.804 0.009** −0.818SO4 0.775 0.014* 0.837Fe tot −0.696 0.037*

⁎ p-value b 0.05.⁎⁎ p-value b 0.01.

All vectors corresponding to metals were collinear and were signifi-cantly associated with the Axis 1 of the MFA (Fig. 10). RV coefficientswere almost identical, ranging from 0.793 to 0.833 (0.011 b p b 0.005)(Table 2).

MFA analysis (Fig. 10, left panel) reflected the evolution with time ofthe communities using site scores computed with i) bacterial phylogenyat the species level and ii) the environmental data set. In this figure, sam-ples were shown as centroids in-between the projected phylogeny andenvironmental data sets. For the interpretation, as both sets of variablesprojectedmore closely, the adequacy between the community structures(in termof diversity and relative proportions of the taxa) grew alongwiththe geochemical environment of their respective habitat. Initially, sampleD0 displayed a very large distance between both sets of variables. As a re-cord, the initial community was dominated by AMD-related taxa, show-ing low population evenness. The relative inadequacy between the twosets of variables could be interpreted either by the indication of stronglyevolving habitat, or by the presence of a cortege of accompanying speciesthat have no specific relationship with the on-going oxidation processes.Recent evidence has shown that a potentially large portion of bacterial di-versity detected in gradient studies was not contributing to specific eco-system functions, being either dead, in a dormant state, or present asextracellular DNA (Krause et al., 2014). In our case, the open-air remedi-ation cell was potentially exposed to various sources of organisms fromverydifferent types of habitats (soils, river, remediated tailing section, cul-tivated land, waste water, rhizosphere and ocean).

ificantly correlated environmental variables are shown only.

vel 0.05 Species level 0.05

RV

nts p-Values Coefficients p-Values

0.002** −0.873 0.002**0.001** −0.871 0.002**0.016** 0.793 0.011*0.016* 0.798 0.010**0.016* 0.798 0.010**0.014* 0.807 0.009**0.009** 0.833 0.005**0.007** 0.812 0.008**0.005** 0.829 0.006**

Cu Al

red

ox

Zn

FeI

I

Mn

SO

4 K pH

o__Syntrophobacteraleso__Desulfobacteraleso__Myxococcaleso__GCA004p__Chlorobio__GN03o__Ignavibacterialeso__Z20o__YS2c__Deltaproteobacteriac__Gemmatimonadeteso__Spirochaetaleso__PirellulalesBacteria - unclassifiedc__Ignavibacteriao__ZB1o__Acholeplasmatalesc__Mollicutesc__OPB56c__Lentisphaeraep__Spirochaeteso__Opitutalesp__PAUC34fc__Spirochaetesp__Proteobacteriap__Acidobacteriac__PRR-12o__Acidobacterialesc__GN13p__Bacteroideteso__Desulfuromonadaleso__Holophagalesp__ABY1_OD1o__Rhodocyclaleso__Nitrospiraleso__Deinococcaleso__Actinomycetaleso__Chromatialeso__Acidithiobacillaleso__Neisserialeso__Acidimicrobialeso__Pseudomonadaleso__Verrucomicrobialeso__Rhizobialeso__Enterobacterialeso__Pasteurellaleso__Lactobacillaleso__Sphingomonadaleso__Streptophytao__Fusobacterialeso__Coriobacterialesc__S15B-MN24o__Gemellales

-0.5 0 0.5Value

2040

60

Color Key and Histogram

Cou

nt

A

B

C

D

Fig. 11.Heat-map formed by the joined analysis of the bacterial data (at the order level) and the environmental data sets. Contributors of both data sets were selected byMFA analysis andonly those significantly correlatedwith Axes 1 and 2were used here.When unknown orderswere found significant, the higher phylogenetic level (class or phylum) providingmeaningfulinformation was inserted instead. Letters A to D describe four clusters of taxa.

119N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

Flooding of the cell had a strong and rapid impact on communities.These displayed a relatively rapid adjustment to evolving environmen-tal conditions. Distances between the community and environmentaldata sets were minimal at D169. The rebound in the relative numbersof AMD-related bacterial guild members at D225, such asAcidithiobacillus spp. and Sulfobacillus spp., provided a strong indicationof rebounding oxidizing conditions due to lower amounts of incomingsurface water with the concomitant intrusion of oceanwater. This sam-ple reflected the presence of small but indicative sequences, affiliatedwith obligate marine bacteria (Pseudoalteromonas spp., Vibrio spp.,Prochlorococcus spp., Halomonas spp., as well as the marine ζ-Proteobacteria (Zeta) Mariprofundus ferrooxydans).

RV coefficients were computed on the community data set, with theselection of taxa that were significantly correlated with axes 1 and 2 oftheMFA. Significantly correlated taxa at the order level were associatedwith significantly correlated environmental variables in a heat-map(Fig. 11). Clustering of the taxa formed two distinct groups, composedof orders significantly correlated with either the initial acidophilic hab-itat or the habitat undergoing the remediation process at circum-neutral pH conditions, reinforcing the role of this latter environmentalvariable in the forging of the community composition (Kuang et al.,2013; Liu et al., 2014). Interestingly, clustering at different phylogeneticlevels provided the same sharp structuring (data not shown), empha-sizing once more the coherence between the bacterial taxa distributionand the ecosystem functioning (Krause et al., 2014). Fig. 11 showed that

the initial acidophilic and aerobic conditions, linked with a cluster ofmetals (Cu, Al, Zn, FeII) were associated with two main clusters of or-ganisms (Clusters A and B), typically found in AMD. Cluster B was ap-parently linked less acidic conditions and with lower concentrations ofMn and sulfate, indicating the presence of taxa linked significantlywith the beginning of the remediation process.

Two main clusters of organisms were associated with overallcircum-neutral pH conditions. Cluster C was composed of organismsthat demonstrated a rapid development from D19 onwards, whichwere favored at early stages of the remediation process. Among theseOrders, sequences affiliated with Desulfuromonadales made the largestcontributions of ca. 10.1% at D77. A high contributions of unknown se-quences, affiliated at the phylum level with Bacteroidetes (25.9% atD47), were found equally in the early stage. Both taxa members gradu-ally decline with time, making only a minor contribution at D225. Clus-ter D included taxa that were shown already to play a major role in theremediation of habitats impacted by AMD (Desulfobacterales andSyntrophobacterales for instance). However, themajority of the taxa as-sociated with this cluster were composed of members that were notlinked specifically with metal and sulfate reduction activities. In thissense, the presence of significantly correlated taxa was interpreted asthe results of an increasing complexity, combinedwith the loading of al-lochthonous populations through immigration, contributing potentiallyto the stabilization of the ecological system through the building-up ofnew niches (Wagg et al., 2014).

120 N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

4. Conclusions

AMD formation caused by marine deposition of copper porphyrysulfide-rich mine tailings forms a source of metal contamination inmarine environments. Their remediation forms a major challenge.In this study, the construction of a wetland on the former oxidationzone considerably reduced sulfide oxidation and created conditionsthat contributed in fine to metal removal from solution. pH andredox were key parameters, which influenced dissolution, sorptionand precipitation mechanisms, and therefore metal immobilizationprocesses. K originated mainly from the allochthonous sources of or-ganic matter, fueling heterotrophic activities, which in turn ensuredproper anoxic growth conditions for both metal and sulfur reducers.In such a system, the development of dedicated flora and the conse-quent accumulation of organic matter could be considered as a ne-cessity, ensuring the accumulation of organic matter that would actas a buffer against oxidizing conditions in case of drought.

Clustering of massive sequencing data at different phylogeneticlevels provided identical responses, illustrating the strong impactcaused by the remediation process on the initial bacterial communi-ty. Significant modifications occurred at higher phylogenetic levels,illustrating niche differentiation between deeply diverging phyloge-netic lineages. This result illustrated the coherence between highbacterial taxa distribution and ecosystem functioning. The simulta-neous presence of both obligate SRB and strictly aerobic organismsfrom D13 on showed that both anaerobic and aerobic activitieswere conducted possibly at the same time. At D225, Sulfobacillusspp. composed ca. 0.3% of the sequences. Less than 0.1% of sequenceswere affiliated with Leptospirillum spp. and Acidithiobacillus spp. Thepresence of these taxa by the end of this study, and possibly theirlong term persistence stress the fragility of the ecosystem and thestrong dependency in water saturation. Large contributions of or-ganisms issued from very different habitats were found in samplescollected in the later stages of this study, although their representa-tiveness could have been underestimated. The positive influence of anet increase in diversity on ecosystem functioning and stability is acommon feature in ecology. In the present study, an increase of thediversity was likely expressed by the development of key ecosystemprocesses that would in turn contribute to the preservation of thehabitat. For instance, neutrophilic photosynthetic bacteria generatednet alkalinity by consuming a weak base (bicarbonate) and produc-ing a strong base (hydroxyl ions), reinforcing the remediation pro-cess. Finally, outcomes of AMD remediation are rarely consideredas a resource per se, through which rich habitats could be created.In this case, the constant flooding of the tailings allowed the makingof a remarkable piece of ecosystem in one of the driest desert onEarth.

Acknowledgments

We would like to thank R. Vicetti and E. Buselli from the SPCC (Ilo,Peru) for making this study possible. We gratefully acknowledge thesupport of N. Espinoza, C. Alfaro, H. Acuña, and the entire staff of theServicios Ambientales, SPCC, for their support during the sampling,field work, and laboratory analysis. We also acknowledge the peoplefrom the LBE (EPFL, Lausanne, Switzerland), the staff of the CAM(University of Lausanne, Switzerland) for their steady and generous as-sistance as well as Dr. R. Flynn (Queen's University Belfast) for the En-glish revision of the manuscript. Thanks to H.-R. Pfeifer, J.-C. Lavanchyand K. Hallberg for their constant support and discussion. Thanks alsoto the people of MATEC working on the Ite Bay tailings remediation.Finally, we would like to express our gratitude to four anonymous re-viewers for their comments which contributed to improve the qualityof this manuscript. This study was financed partially by SPCC inthe framework of his collaboration with the University of Lausanne,Switzerland.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.scitotenv.2015.06.076.

References

Alazard, D., Joseph, M., Battaglia-Brunet, F., Cayol, J.-L., Ollivier, B., 2010. Desulfosporosinusacidiphilus sp. nov.: a moderately acidophilic sulfate-reducing bacterium isolatedfrom acid mining drainage sediments. Extremophiles 14, 305–312.

Baker, B.J., Banfield, J.F., 2003. Microbial communities in acid mine drainage. FEMSMicrobiol. Ecol. 44, 139–152.

Borcard, D., Gillet, F., Legendre, P., 2011. Numerical ecology with R, Use R. Springer Sci.Media http://dx.doi.org/10.1007/978-1-4419-7976-6_1.

Burns, A.S., Pugh, C.W., Segid, Y.T., Behum, P.T., Lefticariu, L., Bender, K.S., 2011. Perfor-mance and microbial community dynamics of a sulfate-reducing bioreactor treatingcoal generated acid mine drainage. Biodegradation 23, 415–429.

Cai, Y., Sun, Y., 2011. ESPRIT-Tree: hierarchical clustering analysis of millions of 16S rRNApyrosequences in quasilinear computational time. Nucleic Acids Res. 39, e95.

Chen, L., Li, J., Chen, Y., Huang, L., Hua, Z., Hu, M., Shu, W., 2013. Shifts in microbial com-munity composition and function in the acidification of a lead/zinc mine tailings.Environ. Microbiol. 15, 2431–2444.

Coates, J.D., Ellis, D.J., Gaw, C.V., Lovley, D.R., 1999. Geothrix fermentans gen. nov., sp. nov., anovel Fe(III)-reducing bacterium from a hydrocarbon-contaminated aquifer. Int.J. Syst. Evol. Microbiol. 49, 1615–1622.

Dean, A.P., Lynch, S., Rowland, P., Toft, B.D., Pittman, J.K., White, K.N., 2013. Natural wet-lands are efficient at providing long-term metal remediation of freshwater systemspolluted by acid mine drainage. Environ. Sci. Technol. 47, 12029–12036.

Dell'Anno, A., Beolchini, F., Rocchetti, L., Luna, G.M., Danovaro, R., 2012. High bacterial bio-diversity increases degradation performance of hydrocarbons during bioremediationof contaminated harbor marine sediments. Environ. Pollut. 167, 85–92.

DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi,D., Hu, P., Andersen, G.L., 2006. Greengenes, a chimera-checked 16S rRNA gene data-base and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072.

Diaby, N., 2009. Biogeochemical Evolution of a Marine Shore Tailings Deposit DuringBioremediation (Thesis University of Lausanne, Lausanne. 217 pp. serval:BIB_5C2714688A50).

Diaby, N., Dold, B., 2014. Evolution of geochemical and mineralogical parameters duringin situ remediation of a marine shore tailings deposit by the implementation of awetland cover. Minerals 4, 578–602.

Dold, B., 2006. Element flows associated with marine shore mine tailings deposits.Environ. Sci. Technol. 40, 752–758.

Dold, B., 2014. Evolution of acidmine drainage in sulfidic mine tailings— a review. Specialissue: mine waste characterization, management and remediation. Minerals 4 (2),621–641.

Dold, B., Fontboté, L., 2001. Element cycling and secondary mineralogy in porphyry cop-per tailings as a function of climate, primary mineralogy, and mineral processing.J. Geochem. Explor. 74, 3–55.

Dold, B., Weibel, L., 2013. Biogeometallurgical pre-mining characterization of oredeposits: an approach to increase sustainability in the mining process. Environ. Sci.Pollut. Res. 20, 7777–7786.

Dold, B., Diaby, N., Spangenberg, J.E., 2011. Remediation of a marine shore tailings depositand the importance of water–rock interaction on element cycling in the coastalaquifer. Environ. Sci. Technol. 45, 4876–4883.

Edgar, R.C., Haas, B.J., Clemente, J.C., Quince, C., Knight, R., 2011. UCHIME improves sensi-tivity and speed of chimera detection. Bioinformatics 27, 2194–2200.

Elshahed, M.S., Senko, J.M., Najar, F.Z., Kenton, S.M., Roe, B.A., Dewers, T.A., Spear, J.R.,Krumholz, L.R., 2003. Bacterial diversity and sulfur cycling in a mesophilic sulfide-rich spring. Appl. Environ. Microbiol. 69, 5609–5621.

Field, D., Tiwari, B., Booth, T., Houten, S., Swan, D., Bertrand, N., Thurston, M., 2006. Opensoftware for biologists: from famine to feast. Nat. Biotechnol. 24, 801–803.

Fierer, N., Bradford, M.A., Jackson, R.B., 2007. Toward an ecological classification of soilbacteria. Ecology 88, 1354–1364.

García-Moyano, A., González-Toril, E., Aguilera, Á., Amils, R., 2012. Comparative microbialecology study of the sediments and the water column of the Río Tinto, an extremeacidic environment. FEMS Microbiol. Ecol. 81, 303–314.

Giloteaux, L., Holmes, D.E., Williams, K.H., Wrighton, K.C., Wilkins, M.J., Montgomery, A.P.,Smith, J.A., Orellana, R., Thompson, C.A., Roper, T.J., Long, P.E., Lovley, D.R., 2013. Char-acterization and transcription of arsenic respiration and resistance genes during insitu uranium bioremediation. ISME 7, 370–383.

Hartman, W.H., Richardson, C.J., Vilgalys, R., Bruland, G.L., 2008. Environmental andanthropogenic controls over bacterial communities in wetland soils. Proc. Natl.Acad. Sci. 105, 17842–17847.

Hua, Z.-S., Han, Y.-J., Chen, L.-X., Liu, J., Hu, M., Li, S.-J., Kuang, J.-L., Chain, P.S., Huang, L.-N.,Shu, W.-S., 2014. Ecological roles of dominant and rare prokaryotes in acid minedrainage revealed by metagenomics and metatranscriptomics. ISME J. http://dx.doi.org/10.1038/ismej.2014.212.

Hughes, J.B., Hellmann, J.J., Ricketts, T.H., Bohannan, B.J.M., 2001. Counting the uncount-able: statistical approaches to estimating microbial diversity. Appl. Environ.Microbiol. 67, 4399–4406.

Huse, S.M., Welch, D.M., Morrison, H.G., Sogin, M.L., 2010. Ironing out the wrinkles in therare biosphere through improved OTU clustering. Environ. Microbiol. 12, 1889–1898.

Johnson, D.B., 2012. Geomicrobiology of extremely acidic subsurface environments. FEMSMicrobiol. Ecol. 81, 2–12.

121N. Diaby et al. / Science of the Total Environment 533 (2015) 110–121

Johnson, D.B., Hallberg, K.B., 2003. The microbiology of acidic mine waters. Res. Microbiol.154, 466–473.

Johnson, D.B., Hallberg, K.B., 2005. Acid mine drainage remediation options: a review. Sci.Total Environ. 338, 3–14.

Johnson, D.B., Hallberg, K.B., 2009. Carbon, iron and sulfur metabolism in acidophilicmicro-organisms. In: Poole, R.K. (Ed.), Advances in Microbial Physiology vol. 54.Academic Press Ltd-Elsevier Science Ltd, London, pp. 201–255.

Jünemann, S., Prior, K., Szczepanowski, R., Harks, I., Ehmke, B., Goesmann, A., Stoye, J.,Harmsen, D., 2012. Bacterial community shift in treated periodontitis patientsrevealed by Ion Torrent 16S rRNA gene amplicon sequencing. PLoS ONE 7, e41606.

Karnachuk, O.V., Gerasimchuk, A.L., Banks, D., Frengstad, B., Stykon, G.A., Tikhonova, Z.L.,Kaksonen, A., Puhakka, J., Yanenko, A.S., Pimenov, N.V., 2009. Bacteria of the sulfurcycle in the sediments of gold mine tailings, Kuznetsk Basin, Russia. Microbiology78, 483–491.

Kaspari, M., Garcia, M.N., Harms, K.E., Santana, M., Wright, S.J., Yavitt, J.B., 2008. Multiplenutrients limit litterfall and decomposition in a tropical forest. Ecol. Lett. 11, 35–43.

Kimura, S., Bryan, C.G., Hallberg, K.B., Johnson, D.B., 2011. Biodiversity and geochemistryof an extremely acidic, low-temperature subterranean environment sustained bychemolithotrophy. Environ. Microbiol. 13, 2092–2104.

Korehi, H., Blöthe, M., Schippers, A., 2014. Microbial diversity at the moderate acidic stagein three different sulfidic mine tailings dumps generating acid mine drainage. Res.Microbiol. 165, 713–718.

Koschorreck, M., Geller, W., Neu, T., Kleinsteuber, S., Kunze, T., Trosiener, A., Wendt-Potthoff, K., 2010. Structure and function of the microbial community in an in situreactor to treat an acidic mine pit lake. FEMS Microbiol. Ecol. 73, 385–395.

Krause, S., Le Roux, X., Niklaus, P.A., Van Bodegom, P.M., Lennon, J.T., Bertilsson, S.,Grossart, H.-P., Philippot, L., Bodelier, P.L.E., 2014. Trait-based approaches for under-standing microbial biodiversity and ecosystem functioning. Aquat. Microbiol. 5, 251.

Kuang, J.-L., Huang, L.-N., Chen, L.-X., Hua, Z.-S., Li, S.-J., Hu, M., Li, J.-T., Shu, W.-S., 2013.Contemporary environmental variation determines microbial diversity patterns inacid mine drainage. ISME J. 7, 1038–1050.

Liu, J., Hua, Z.-S., Chen, L.-X., Kuang, J.-L., Li, S.-J., Shu,W.-S., Huang, L.-N., 2014. Correlatingmicrobial diversity patterns with geochemistry in an extreme and heterogeneous en-vironment of mine tailings. Appl. Environ. Microbiol. 80, 3677–3686.

Lovley, D.D., 2013. Dissimilatory Fe(III)- andMn(IV)-reducing prokaryotes. In: Rosenberg,E., DeLong, E.F., Lory, S., Stackebrandt, E., Thompson, F. (Eds.), The Prokaryotes.Springer Berlin Heidelberg, pp. 287–308.

Mayes, W.M., Batty, L.C., Younger, P.L., Jarvis, A.P., Kõiv, M., Vohla, C., Mander, U., 2009.Wetland treatment at extremes of pH: a review. Sci. Total Environ. 407, 3944–3957.

McLean, J.S., Lombardo, M.-J., Badger, J.H., et al., 2013. Candidate phylum TM6 genome re-covered from a hospital sink biofilm provides genomic insights into this uncultivatedphylum. Proc. Natl. Acad. Sci. 110, E2390–E2399.

Moreau, J.W., Fournelle, J.H., Banfield, J.F., 2013. Quantifying heavy metals sequestrationby sulfate-reducing bacteria in an acid mine drainage-contaminated natural wetland.Front. Microbiol. http://dx.doi.org/10.3389/fmicb.2013.00043.

Moro, H., Kunito, T., Saito, T., Yaguchi, N., Sato, T., 2014. Soil microorganisms are lesssusceptible than crop plants to potassium deficiency. Arch. Agron. Soil Sci. 60,1807–1813.

Ňancucheo, I., Johnson, D.B., 2012. Selective removal of transition metals from acidicminewaters by novel consortia of acidophilic sulfidogenic bacteria. Microb. Biotechnol. 5,34–44.

Petrie, L., North, N.N., Dollhopf, S.L., Balkwill, D.L., Kostka, J.E., 2003. Enumeration andcharacterization of iron(III)-reducing microbial communities from acidic subsurfacesediments contaminated with uranium(VI). Appl. Environ. Microbiol. 69, 7467–7479.

Philippot, L., Andersson, S.G.E., Battin, T.J., Prosser, J.I., Schimel, J.P., Whitman, W.B., Hallin,S., 2010. The ecological coherence of high bacterial taxonomic ranks. Nat. Rev.Microbiol. 8, 523–529.

Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., Glockner, F.O.,2013. The SILVA ribosomal RNA gene database project: improved data processing andweb-based tools. Nucleic Acids Res. 41, D590–D596.

R Development Core Team, 2009. R: a language and environment for statistical comput-ing. http://www.R-project.org.

Sánchez-Andrea, I., Stams, A.J.M., Amils, R., Sanz, J.L., 2013. Enrichment and isolation of ac-idophilic sulfate-reducing bacteria from Tinto River sediments. Environ. Microbiol.Rep. 5 (5), 672–678.

Sánchez-Andrea, I., Sanz, J.L., Bijmans, M.F.M., Stams, A.J.M., 2014. Sulfate reduction at lowpH to remediate acid mine drainage. J. Hazard. Mater. 269, 98–109.

Schloss, P.D., Westcott, S.L., Ryabin, T., et al., 2009. Introducing mothur: open-source,platform-independent, community-supported software for describing and compar-ing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541.

Sheoran, A.S., Sheoran, V., 2006. Heavy metal removal mechanism of acid mine drainagein wetlands: a critical review. Miner. Eng. 19, 105–116.

Smuda, J., Dold, B., Spangenberg, J.E., Bustos, C., Kobek, M., Pfeifer, H.R., 2014. Element cy-cling during the transition from alkaline to acidic environment in an active porphyrycopper tailings impoundment, Chuquicamata, Chile. J. Geochem. Explor. 140, 23–40.

Sogin, M.L., Morrison, H.G., Huber, J.A., Welch, D.M., Huse, S.M., Neal, P.R., Arrieta, J.M.,Herndl, G.J., 2006. Microbial diversity in the deep sea and the underexplored “rarebiosphere.”. Proc. Natl. Acad. Sci. 103, 12115–12120.

Volant, A., Bruneel, O., Desoeuvre, A., et al., 2014. Diversity and spatiotemporal dynamicsof bacterial communities: physicochemical and other drivers along an acid minedrainage. FEMS Microbiol. Ecol. 90, 247–263.

Wagg, C., Bender, S.F., Widmer, F., van der Heijden, M.G.A., 2014. Soil biodiversity and soilcommunity composition determine ecosystem multifunctionality. Proc. Natl. Acad.Sci. 111, 5266–5270.

Zavarzina, D.G., Sokolova, T.G., Tourova, T.P., Chernyh, N.A., Kostrikina, N.A., Bonch-Osmolovskaya, E.A., 2007. Thermincola ferriacetica sp. nov., a new anaerobic, thermo-philic, facultatively chemolithoautotrophic bacterium capable of dissimilatory Fe(III)reduction. Extrem. Life Extreme Cond. 11, 1–7.

Ziegler, S., Dolch, K., Geiger, K., et al., 2013. Oxygen-dependent niche formation of apyrite-dependent acidophilic consortium built by archaea and bacteria. ISME J. 7,1725–1737.