**Genomic and Biotechnological Characterization of the Heavy-Metal Resistant, Arsenic-Oxidizing Bacterium** *Ensifer* **sp. M14**

**George C diCenzo 1,\*,†, Klaudia Debiec 2,\*,†, Jan Krzysztoforski 3, Witold Uhrynowski 2, Alessio Mengoni 1, Camilla Fagorzi 1, Adrian Gorecki 4, Lukasz Dziewit 4, Tomasz Bajda 5, Grzegorz Rzepa <sup>5</sup> and Lukasz Drewniak <sup>2</sup>**


Received: 14 June 2018; Accepted: 25 July 2018; Published: 27 July 2018

**Abstract:** *Ensifer* (*Sinorhizobium*) sp. M14 is an efficient arsenic-oxidizing bacterium (AOB) that displays high resistance to numerous metals and various stressors. Here, we report the draft genome sequence and genome-guided characterization of *Ensifer* sp. M14, and we describe a pilot-scale installation applying the M14 strain for remediation of arsenic-contaminated waters. The M14 genome contains 6874 protein coding sequences, including hundreds not found in related strains. Nearly all unique genes that are associated with metal resistance and arsenic oxidation are localized within the pSinA and pSinB megaplasmids. Comparative genomics revealed that multiple copies of high-affinity phosphate transport systems are common in AOBs, possibly as an As-resistance mechanism. Genome and antibiotic sensitivity analyses further suggested that the use of *Ensifer* sp. M14 in biotechnology does not pose serious biosafety risks. Therefore, a novel two-stage installation for remediation of arsenic-contaminated waters was developed. It consists of a microbiological module, where M14 oxidizes As(III) to As(V) ion, followed by an adsorption module for As(V) removal using granulated bog iron ores. During a 40-day pilot-scale test in an abandoned gold mine in Zloty Stok (Poland), water leaving the microbiological module generally contained trace amounts of As(III), and dramatic decreases in total arsenic concentrations were observed after passage through the adsorption module. These results demonstrate the usefulness of *Ensifer* sp. M14 in arsenic removal performed in environmental settings.

**Keywords:** *Ensifer* (*Sinorhizobium*) sp. M14; arsenic-oxidizing bacteria; heavy metal resistance; draft genome sequence; comparative genomic analysis; biosafety; biotechnology for arsenic removal; adsorption; water treatment; in situ (bio)remediation

### **1. Introduction**

The development and implementation of bioremediation technologies based on bioaugmentation requires the selection of appropriate microbial strains. A basic requirement of strains used as bioaugmentation agents is their ability to survive in the environment into which they are introduced. Thus, such strains are usually characterized by high tolerance to heavy metals [1,2], resistance and ability to use organic (sometimes toxic) compounds [3,4], resistance to antibiotics [5], and an ability to thrive in the presence of local bacteriophages and microorganisms. Another important feature of strains used in bioaugmentation is their ability to perform effective transformation of the particular compound under changing environmental conditions (e.g., temperature, humidity, and pH). This is always the critical limitation, as many strains effective under laboratory conditions are, in fact, ineffective in field applications. Microorganisms suitable in bioremediation should maintain their activity in various seasons and under variable substrate inflow. A very important factor influencing the decision to apply a given microorganism in practice is also its interaction with the environment [6]. Strains that contribute to the uncontrolled release of contaminants, dissemination of antibiotic resistance genes, or disrupt the functioning of the ecosystem (e.g., by eliminating key microorganisms) should not be applied in open (uncontrolled) usage.

In this study, we provide a detailed characterization of *Sinorhizobium* sp. M14 (renamed here to *Ensifer* sp. M14 due to its phylogenetic positioning within the *Ensifer* clade), which is a strain with high potential to be used in bioremediation technologies for the removal of arsenic from contaminated waters and wastewaters. *Ensifer* sp. M14 is a psychrotolerant strain that was isolated from the microbial mats present in the arsenic-rich bottom sediments of an abandoned gold mine in Zloty Stok (Poland) [7]. The arsenic concentration in the mine waters reaches ~6 mg L<sup>−</sup>1, while in the microbial mats the level of accumulated arsenic is close to 20 g L−<sup>1</sup> [8]. Previous physiological studies showed that *Ensifer* sp. M14 tolerates extremely high concentrations of arsenate [As(V)—up to 250 mM] and arsenite [As(III)—up to 20 mM], and is able to oxidize As(III) both chemolithoautotrophically [using arsenite or arsenopyrite (FeAsS) as a source of energy] and heterotrophically [7]. Batch experiments performed under various conditions of pH, temperature, and arsenic concentration confirmed the high adaptive potential of *Ensifer* sp. M14 [9]. The strain was capable of intensive growth and efficient biooxidation in a wide range of conditions, including low temperature [As(III) oxidation rate = 0.533 mg L−<sup>1</sup> h−<sup>1</sup> at 10 ◦C]. Continuous flow experiments under environment-like conditions (2 L flow bioreactor) showed that *Ensifer* sp. M14 efficiently transforms As(III) into As(V) [24 h of residence time was sufficient to oxidize 5 mg L−<sup>1</sup> of As(III)], but its activity depended mainly on the retention time in the bioreactor, which may be accelerated by stimulation with yeast extract as a source of nutrients [9].

Analysis of the extrachromosomal replicons of *Ensifer* sp. M14 revealed that its arsenic metabolism properties are linked with the presence of the mega-sized plasmid pSinA (109 kbp) [10]. The loss of the pSinA plasmid from *Ensifer* sp. M14 cells (using a target-oriented replicon curing technique [11]) eliminated the ability to oxidize As(III), and caused deficiencies in resistance to arsenic and heavy metals (Cd, Co, Zn, and Hg). In turn, the introduction of this plasmid into other representatives of the *Alphaproteobacteria* showed that cells with pSinA acquired the ability to oxidize arsenite and exhibited higher tolerance to arsenite than their parental, pSinA-less, wild-type strains. Horizontal transfer of arsenic metabolism genes by *Ensifer* sp. M14 was also confirmed in microcosm experiments [10]. The plasmid pSinA was successfully transferred via conjugation into indigenous bacteria of *Alpha*- and *Gammaproteobacteria* classes from the microbial community of As-contaminated soils. Transconjugants carrying plasmid pSinA expressed arsenite oxidase and stably maintained pSinA in their cells after approximately 60 generations of growth under nonselective conditions [10].

The second mega-sized replicon of *Ensifer* sp. M14—plasmid pSinB (300 kbp)—also plays an important role in the adaptation of the host to the mine environment. Structural and functional analysis of this plasmid showed that it carries gene clusters involved in heavy metals resistance. Among these are genes encoding efflux pumps, permeases, transporters, and copper oxidases, which are responsible for resistance to arsenic, cobalt, zinc, cadmium, iron, mercury, nickel, copper, and silver [12].

In this paper, we obtained a draft genomic sequence of *Ensifer* sp. M14 and performed complex genome-guided characterization of this bacterium. Special considerations were given to (i) determination of the metabolism of phosphate, sulfur, iron, and one-carbon substrates, and (ii) investigation of the biosafety of *Ensifer* sp. M14 in the context of its release to the environment (e.g., determination of the presence of virulence and antibiotic resistance genes). These analyses revealed hints about the potential application of this strain in biotechnological applications; for example, the ability of it to survive environmental stresses, and whether it is likely to pose a safety risk. As the genomic analyses were consistent with *Ensifer* sp. M14 having potential application in biotechnology, we performed a large-scale simulation of the usage of M14 in the biological and chemical removal of arsenic from contaminated waters. The results support that the developed low-cost approach is an efficient method for the removal of arsenic from contaminated water.

### **2. Materials and Methods**

### *2.1. Genome Sequencing, Assembly, and Annotation*

*Ensifer* sp. M14 (available on request from the authors) was grown at 30 ◦C to stationary phase in TY medium (5 g L−<sup>1</sup> tryptone, 3 g L−<sup>1</sup> yeast extract, and 0.4 g L−<sup>1</sup> calcium chloride). Genomic DNA was isolated from the culture using a cetyltrimethylammonium bromide (CTAB) method [13] modified for bacterial DNA isolation as described by the Joint Genome Institute [14]. Sequencing was performed at IGATech (Udine, Italy) using an Illumina HiSeq2500 instrument with 125-bp paired-end reads. Two independent sequencing runs were performed. Reads were assembled into scaffolds using SPAdes v3.9.0 [15,16]. The scaffolds returned by SPAdes were parsed to remove those with less than 10× coverage or with a length below 200 nucleotides. Using FastANI [17], one-way average nucleotide identity (ANI) of the *Ensifer* sp. M14 assembly was calculated against the 887 alpha-proteobacterial genomes available through the National Center for Biotechnological Information (NCBI) with an assembly level of 'complete' or 'chromosome'. The 10 genomes most closely related to *Ensifer* sp. M14 were identified on the basis of the ANI results. These 10 genomes, together with the complete pSinA and pSinB plasmid sequences [10,12], were used as reference genomes for further scaffolding of the assembly using MeDuSa [18]. The *Ensifer* sp. M14 assembly was then annotated using prokka version 1.12-beta [19], annotating coding regions with Prodigal [20], tRNA with Aragon [21], rRNA with Barrnap (github.com/tseemann/barrnap), and ncRNA with Infernal [22] and Rfam [23]. The predicted coding sequences were associated with Cluster of Orthologous Genes (COG) categories, Gene Ontology (GO) terms, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway terms, and eggNOG annotations using eggNOG-mapper version 0.99.2-3-g41823b2 [24]. The assembly was deposited to NCBI with the GenBank accession QJNR00000000 (the version described in this paper is version QJNR01000000) and the BioSample accession SAMN09254189.

### *2.2. Phylogenetic Analysis*

Initially, all 133 *Sinorhizobium*/*Ensifer* genomes available through NCBI, regardless of assembly level, were downloaded. FastANI [17] was used to calculate one-way ANI values between *Ensifer* sp. M14 and each of the 133 downloaded genomes. Only the strains meeting at least one of the following two requirements were kept for further analyses: (i) had a genome assembly level of 'complete' or 'chromosome', or (ii) had an ANI value of at least 85% compared to *Ensifer* sp. M14. This resulted in a final set of 46 strains, when including *Ensifer* sp. M14.

The pangenome of the 46 strains was calculated using Roary version 3.11.3 [25], as described below, following re-annotation with prokka version 1.12-beta [19]. Included in the Roary output was a concatenated nucleotide alignment of the 1652 core genes, each individually aligned with PRANK [26]. The core gene alignment was used to build a maximum likelihood phylogeny with RAxML version 8.2.9 [27] using the following command:

raxmlHPC-HYBRID-SSE3-T 5-s input.fasta-N autoMRE-n output-f a-p 12345-x 12345-m GTRCAT.

The final tree is the bootstrap best tree following 50 bootstrap replicates, and was visualized using the online iTOL (Interactive Tree of Life) webserver [28].

Strains were grouped into putative species on the basis of ANI and average amino acid identity (AAI) values, using thresholds of 96% for both measures. Groupings for ANI were the same at thresholds of 96% and 94%. Pairwise ANI values were calculated between each strain using FastANI [17], and the values in both directions were averaged. The CompareM workflow (github.com/dparks1134/CompareM) was used for calculating the AAI values. In the CompareM workflow, orthologous proteins were first identified using DIAMOND with the sensitive setting [29], and thresholds of 40% identity over 70% the length of the protein and a maximum e-value of 1e−<sup>12</sup> were applied, as these are the thresholds used in the myTaxa program [30].

### *2.3. Sinorhizobium/Ensifer Pangenome Calculation*

All 46 strains included in the phylogenetic analyses were reannotated using prokka version 1.12-beta [19], to ensure consistent annotation. The pangenome of the 46 reannotated strains was then determined with Roary version 3.11.3 [25], using an amino acid identity threshold of 80% and the following command:

roary-p 20-f Output-e-I 80-g 100,000 Input/\*.gff.

For comparison of the gene content of *Ensifer* sp. M14, *Ensifer* sp. A49, *Ensifer adhaerens* OV14, and *Ensifer adhaerens* Casida A, the data was extracted from the full 46-strain pangenome. The complete gene presence/absence output from Roary is provided as Data Set S1. Several short proteins of *Ensifer* sp. M14 were not present in the output of the Roary analysis; these proteins were not considered when identifying unique genes.

### *2.4. Comparative Genomics of Arsenic Oxidizing Bacteria*

The genomes of *Agrobacterium tumefaciens* 5A [31], *Agrobacterium tumefaciens* Ach5 [32], *Ensifer adhaerens* OV14 [33], *Neorhizobium galegae* HAMBI 540 [34], and *Rhizobium* sp. NT-26 [35] were downloaded from NCBI GenBank and reannotated using prokka, as described above for *Ensifer* sp. M14. The GenBank files of the re-annotated genomes, and the *Ensifer* sp. M14 genome, were uploaded to the KBase webserver [36], and OrthoMCL [37] was run on the KBase server using an e-value threshold of 1e−12. Identification of phosphate transport and arsenic resistance genes in other bacterial genomes (*Achromobacter arsenitoxydans* SY8 [38], *Herminiimonas arsenicoxydans* ULPAs1 [39], and *Pseudomonas stutzeri* TS44 [40]) was accomplished by manually searching the GenBank file of the RefSeq annotated genomes [41].

### *2.5. Identification of Prophage Loci*

PhiSpy version 3.2 [42], implemented in Python, was used to predict phage genes. The *Ensifer* sp. M14 GenBank file produced with prokka was converted to SEED format using the genbank\_to\_seed.py script. The converted file was then used as input for the PhiSpy.py script, using the generic test set for training.

### *2.6. Identification of Putative Antibiotic Resistance Genes*

To identify putative antibiotic resistance genes, the Resistance Gene Identifier (RGI) in the Comprehensive Antibiotic Resistance Database (CARD) software was used [43]. Hits showing at least 50% identity with the reference protein were considered significant. Each hit was verified manually using BLASTp analysis.

### *2.7. Analysis of the Antimicrobial Susceptibility Patterns*

To determine the antimicrobial susceptibility patterns of *Ensifer* sp. M14, minimum inhibitory concentrations (MICs) of 11 antimicrobial agents were assessed using Etest™ (Liofilchem, Roseto degli Abruzzi, Italy). The analysis was conducted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) recommendations [44]. The following antibiotics (selected based on the bioinformatic analyses that identified putative antibiotic resistance genes) were used: (i) aminoglycosides–gentamicin (GN; concentration of antibiotic: 0.064–1024 μg mL−<sup>1</sup> Roseto degli Abruzzi1); (ii) β-lactams (penicillin derivatives)–ampicillin (AMP; 0.016–256 μg mL<sup>−</sup>1); (iii) β-lactams (cephalosporins)–cefixime (CFM; 0.016–256 μg mL−1); (iv) β-lactams (cephalosporins)–cefotaxime (CTX; 0.016–256 μg mL−1); (v) β-lactams (cephalosporins)–ceftriaxone (CRO; 0.016–256 μg mL−1); (vi) fluroquinolones–ciprofloxacin (CIP; 0.002–32 μg mL−1); (vii) fluroquinolones–moxifloxacin (MXF; 0.002–32 μg mL−1); (viii) phenicols–chloramphenicol (C; 0.016–256 μg mL−1); (ix) rifamicyns–rifampicin (RD; 0.016–256 μg mL−1); (x) sulfonamides–trimethoprim (TM; 0.002–32 μg mL−1); and (xi) tetracyclines–tetracycline (TE; 0.016–256 μg mL−1). The susceptibility testing was performed at 30 ◦C for 20 h. After incubation, plates were photographed and MICs were defined. Antimicrobial susceptibility data were interpreted according to the EUCAST breakpoint table version 8.0 [45].

### *2.8. Search for Symbiotic Proteins*

A custom pipeline based on the use of hidden Markov models (HMM) was used to search the proteomes of all 46 *Sinorhizobium*/*Ensifer* strains for the presence of the nodulation proteins NodA, NodB, and NodC, as well as for the nitrogenase proteins NifH, NifD, and NifK. This pipeline is dependent on HMMER version 3.1b2 [46], and the complete Pfam-A version 31.0 (16,712 HMMs) and TIGERFAM version 15.0 (4488 HMMs) databases [47,48]. After downloading the HMM databases, hmmconvert was used to ensure consistent formatting. The two databases were combined into a single HMM database, and then converted into a searchable database with hmmpress. Additionally, the HMM seed alignments for NodA (TIGR04245), NodB (TIGR04243), NodC (TIGR04242), NifH (TIGR01287), NifD (TIGR01282), and NifK (TIGR01286) were downloaded from the TIGRFAM database [47].

For each HMM seed alignment, a HMM was built using hmmbuild, and the output was then searched against the complete set of *Sinorhizobium*/*Ensifer* proteins using hmmsearch. The output was parsed, and the amino acid sequences for each of the hits (regardless of e-value) were collected. Each set of sequences were then searched against the combined HMM database using hmmscan, and the output parsed to identify the top scoring HMM hit for each query protein. Proteins were annotated as follows: NodA if the top hit was TIGR04245 (TIGRFAM) or NodA (Pfam); NodB if the top hit was TIGR04243 (TIGRFAM); NodC if the top hit was TIGR04242 (TIGRFAM); NifH if the top hit was TIGR01287 (TIGRFAM) or Fer4\_NifH (Pfam); NifD if the top hit was TIGR01282 (TIGRFAM), TIGR01860 (TIGRFAM), or TIGR01861 (TIGRFAM); NifK if the top hit was TIGR02932 (TIGRFAM), TIGR02931 (TIGRFAM), or TIGR01286 (TIGRFAM).

### *2.9. Cluster of Orthologous Genes Functional Annotation*

Proteomes were annotated with COG functional categories using eggNOG-mapper version 0.99.2-3-g41823b2 [24]. The output of eggNOG-mapper was parsed with a custom Perl script to count the percentage of proteins annotated with each functional category. Fisher exact tests, performed using MATLAB R2016b (www.mathworks.com), were performed to identify statistically significant differences (*p* < 0.05) between *Ensifer* sp. M14 and the other strains.

### *2.10. In Silico Metabolic Reconstruction and Constraint-Based Modelling*

Metabolic reconstruction steps and constraint-based metabolic modeling were performed in MATLAB 2017a (Mathworks, Natick, MA, USA), using the Gurobi 7.0.2 solver (gurobi.com), SBMLToolbox 4.1.0 [49], libSBML 5.13.0 [50], and scripts from the COBRA Toolbox [51] and the Tn-Core Toolbox [52]. The ability of the *Ensifer* sp. M14 model to grow when individually provided with 163 carbon sources was tested using flux balance analysis (FBA) as implemented in the 'optimizeCbModel' function of the COBRA Toolbox.

An initial draft metabolic reconstruction was prepared using the online KBase webserver [36]. The *Ensifer* sp. M14 genome was uploaded and re-annotated with RAST functions using the 'annotate microbial genome' function. The re-annotated genome was used to build a draft model with the 'build

metabolic model' function, performing gap-filling on a glucose minimal medium, and with automatic biomass template selection. This reconstruction was downloaded in SBML format, and then imported into MATLAB as a COBRA formatted metabolic model for further manipulation. After removing duplicate genes from the gene list and updating the gene-reaction rules appropriately, the model was expanded based on the reaction content of the curated iGD1575 and iGD726 metabolic reconstructions of the closely related species *Sinorhizobium meliloti* [53,54]. First, a BLAST bidirectional best hit approach was used to identify putative orthologs (at least 70% identity over at least 70% the protein length) between *S. meliloti* Rm1021 and *Ensifer* sp. M14. All *S. meliloti* genes without a putative ortholog in *Ensifer* sp. M14 were deleted from the iGD1575 and iGD726 models, and the constrained reactions removed. Next, the reactions of iGD726 and the draft *Ensifer* sp. M14 model were compared based on their equations, and all reactions unique to iGD726 were identified and transferred to the *Ensifer* sp. M14 model. Exceptions were iGD726 reactions that differed from a reaction in the *Ensifer* sp. M14 model only in the presence/absence of a proton or in metabolite stoichiometry. This process was then repeated, transferring the unique reactions of iGD1575 to the partially expanded model. When transferring reactions, associated genes were also transferred and changed to the name of the *Ensifer* sp. M14 orthologs. Following the expansions, all reactions producing dead-end metabolites were iteratively removed from the model. The final model contained 1491 genes, 1561 reactions, and 1105 metabolites, and is available in Data Set S2.

### *2.11. Prediction of Secondary Metabolism*

Loci encoding secondary metabolic pathways were predicted in the *Ensifer* sp. M14 genome using the antiSMASH webserver [55]. The *Ensifer* sp. M14 GenBank file was uploaded to the bacterial version of antiSMASH, and the analysis was run with all options selected with default parameters.

### *2.12. Construction of a Pilot-Scale Installation for Arsenic Bioremediation*

A pilot-scale installation for the removal of arsenic from contaminated waters was developed. The installation was operated using water from a dewatering system of a former gold mine located in the Zloty Stok area (SW Poland), which is highly polluted with arsenic. The total arsenic concentration, arsenic speciation, as well as detailed chemical and physical characteristics of the water are presented elsewhere [56]. The installation consisted of two modules: the microbiological module and the adsorption module (Figure 1).

**Figure 1.** The pilot-scale installation used for remediation of arsenic contaminated water. The image is a schematic representation of the pilot-scale installation developed as part of this work. Both the microbiological and adsorption modules are shown.

The microbiological module was based on the activity of *Ensifer* sp. M14, which was used as an arsenite biooxidizer. This module included a 200 L bioreactor with an electric heater. The contaminated water flowing out from the gold mine was fed into the bioreactor through a pressure

reducer and a peristaltic pump at a volume flow rate of 8.33 L h−1, corresponding to a residence time of 24 h in the bioreactor. Outflow of the water occurred as overflow in the upper part of the bioreactor. To increase the effectiveness of the arsenite biooxidition, the bioreactor was equipped with an additional aeration system that consisted of an air pump producing compressed air. The additional aeration system was included in our previous study and showed that the arsenite oxidation efficiency of *Ensifer* sp. M14 is higher in the presence of additional aeration during continuous culturing [9]. Moreover, yeast extract was added to the bioreactor as a source of vitamins (growth supplements). Fifty grams of powdered yeast extract (Sigma-Aldrich, St. Louis, MO, USA) was added to the bioreactor twice a week. This was done as we previously observed that the presence of yeast extract led to an increase in the growth and efficiency of arsenite biooxidation of *Ensifer* sp. M14 during continuous culturing [9]. This relationship was also confirmed in other papers [10,57]. The supply of air also contributed to the mixing of the bioreactor content. The bioreactor was equipped with a multifunctional electrode dedicated to controlling the chemical and physical parameters of water, specifically, to monitor pH, redox potential, and temperature (Hydrolab HL4, OTT Hydromet, Kempten, Germany). The water leaving the bioreactor was fed into a 60 L buffer tank, which functioned as the connecting element between the bioreactor and the adsorption module. The inclusion of the buffer tank helped maintain a constant water level in the adsorption columns and ensured a constant flow of water from the bioreactor to the adsorption columns.

The adsorption module consisted of three columns (17 L volume each) filled with granulated bog iron ores (about 15 kg per column) and connected in series (Figure 1). The detailed chemical and physical parameters, chemical composition, and stability of the adsorbent were presented previously [56]. Contaminated water from the buffer tank (after passing through the microbiological module) was fed into the first column using a second peristaltic pump at a volume flow rate of 8.33 L h<sup>−</sup>1, which corresponded to approximately one hour of residence time per column.

The installation was also equipped with a process control system (operated at the location of the pilot plant or remotely via a Global System for Mobile Communications (GSM)) that monitored and controlled key process parameters including the volume flow rate of the water, the water temperature at the inlet, in the bioreactor, and at the outlet of the pilot plant, as well as the ambient temperature.

### *2.13. Installation Start-Up*

Scale-up of the installation (from laboratory scale to pilot scale) required the development of procedures for successful start-up based on the results of our previous study [9]. The first step of the start-up of the microbiological module was inoculation of the bioreactor with an appropriate amount of *Ensifer* sp. M14. The bioreactor filled with arsenic contaminated water was inoculated with 200 mL of a highly concentrated overnight culture of *Ensifer* sp. M14 suspended in 0.85% NaCl solution. The initial OD600 in the bioreactor was 0.01. In earlier experiments, it was determined that a starting cell density of 10<sup>8</sup> CFU mL−<sup>1</sup> (which corresponds to an OD600 of 0.1) is required for the installation to work properly [9]. To increase the density of the *Ensifer* sp. M14, the water in the bioreactor was supplemented with powdered yeast extract to a final concentration of 0.04%. Additionally, aeration was applied. Finally, the temperature of the water was increased (from 10 to 22 ◦C) with the use of an electric heater placed in the bioreactor. Application of all these treatments led to an OD600 value of 0.1 within 24 h.

Start-up procedures related to the adsorption module mainly concerned the preparation of the adsorbent for its usage. After filling the columns with granulated bog iron ores, it was necessary to condition the adsorbent (rinsing the adsorbent with the tap water without arsenic) to remove all the loosely bound fractions.

### *2.14. Biological and Chemical Analyses*

Arsenic speciation was investigated with the use of ion chromatography on an IonPac AS18 (2 mm, Dionex, Lübeck, Germany) column on an ICS Dionex 3000 (Lübeck, Germany) instrument equipped with an ASRS® 2 mm suppressor, which was coupled to a ZQ 2000 mass spectrometer via an electrospray source (Waters, Milford, MA, USA) according to the method described by Debiec et al. [9]. In the adsorption module, the total arsenic concentration was investigated. Total arsenic concentration was measured using a Graphite Furnace Atomic Absorption Spectrometry (GFAAS; AA Solaar M6 Spectrometer, TJA Solutions, Waltham, MA, USA). Arsenic standard solutions (Merck, Darmstadt, Germany) were prepared in 3% HNO3. The pH and redox potential were measured only in the microbiological module. Samples of raw water, water from the bioreactor, as well as water at the inflow and outflow of each adsorption column were collected once a day during the first 8 days, and then three times a week up to day 40. Samples taken from the bioreactor were stored at −20 ◦C, while samples collected from the adsorption module were stored at 4 ◦C. This experiment was repeated twice.

### **3. Results and Discussion**

### *3.1. Sequencing of the Ensifer sp. M14 Genome*

The draft genome sequence of *Ensifer* sp. M14 was obtained as described in the Materials and Methods, and the general genomic features are described in Table 1.


**Table 1.** Features of the *Ensifer* sp. M14 genome assembly.

\* As determined using eggnog-mapper [24]. Those genes not returned in the eggNOG-mapper output were said to have no similarity; † Excluding those annotated with COG category S (unknown function); ¥ Excluding those annotated as protein/domain of unknown/uncharacterized function. CDS (Coding Sequences); COG (Cluster of Orthologous Genes); KEGG (Kyoto Encyclopedia of Genes and Genomes); GO (Gene Ontology).

The assembly consists of 7,345,249 bp spread over 45 scaffolds at an average coverage of 118×. Of the 45 scaffolds, 12 are over 40 kbp in size and account for 98.7% of the assembly. Based on similarity searches of the scaffolds, previous plasmid profiling of *Ensifer* sp. M14 [10,12], and the finished genomes of related strains [33,58], we predict that the *Ensifer* sp. M14 genome consists of one chromosome (at least 4.4 Mbp in size), two additional large replicons (chromids and/or large megaplasmids, at least 1.6 Mbp and 0.6 Mbp in size), and the two previously reported smaller megaplasmids (pSinA and pSinB, 109 kbp and 300 kbp, respectively, based on previous papers [10,12]). A total of 6874 coding sequences were predicted, which is more than the 6218 predicted in *S. meliloti* Rm1021 and the 6641 of *E. adhaerens* Casida A, but less than the 7033 predicted in *E. adhaerens* OV14 [33,58,59]. Six putative prophages were identified on Scaffold 4 (the chromosome) using PhiSpy [42]; these ranged in size from 21 to 65 genes, and accounted for a total of 292 genes (Data Set S3). However, no CRISPR loci were detected during annotation with prokka [19]; a questionable, short CRIPSR with one spacer was detected with CRISPRfinder [59], but its location within a predicted coding region suggests it is unlikely to be a true CRISPR locus. No evidence for the presence of the common nodulation genes *nodABC* or the nitrogenase genes *nifHDK* was found using a hidden Markov model based approach. The *Ensifer* sp. M14 assembly has been deposited in GenBank under the accession QJNR00000000, as part of the BioSample SAMN09254189.

### *3.2. Taxonomic Analysis of Ensifer sp. M14*

Phylogenetic analyses were performed to identify the relationships between *Ensifer* sp. M14 and previously sequenced *Sinorhizobium*/*Ensifer* strains. Forty-five *Sinorhizobium*/*Ensifer* genomes were downloaded from the NCBI database (see Materials and Methods for criteria for strain inclusion), and a maximum likelihood phylogeny of these strains plus *Ensifer* sp. M14 was built based on 1652 core genes (Figure 2).

The 46 strains were grouped into putative species on the basis of whole genome ANI and AAI values (Figures S1 and S2). The results revealed that *Ensifer* sp. M14 is closely related to *Ensifer* (*Sinorhizobium*) sp. A49 (98.5% ANI and 98.9% AAI), and that these strains likely belong to a new species. *Ensifer* sp. A49 was previously isolated from soil of the Fureneset Rural Development Centre of Fjaler, Norway [60]. However, the pSinA and pSinB plasmids, carrying genes involved in arsenic oxidation and heavy metal resistance [10,12], appear to be specific to *Ensifer* sp. M14 and may therefore have been gained during growth in the Zloty Stok gold mine [7]. The most closely related named species is *Ensifer adhaerens*, which includes bacterial predators capable of feeding on organisms such as *Micrococcus luteus* [58,61].

**Figure 2.** Phylogeny of a selected 46 *Sinorhizobium*/*Ensifer* strains with a publicly available whole genome sequence. An unrooted RAxML maximum likelihood phylogeny of 46 *Sinorhizobium*/*Ensifer* strains was prepared on the basis of the concatenated nucleotide alignments of 1652 core genes. The presented tree is the bootstrap best tree following 50 bootstrap replicates, and the scale represents the mean number of nucleotide substitutions per site. Nodes with 100% bootstrap support are indicated by the black circles. The colors and numbers to the right of the tree are used to indicate strains that group into putative species on the basis of average nucleotide identity (>96% ANI; same results were obtained with >94% ANI) and average amino acid identity (>96% AAI), as described in the Materials and Methods. Type strains are indicated by the 'T'. The accessions for all strains included in this figure are provided in Table S1.

### *3.3. Identification of Unique Features of the Ensifer sp. M14 Genome*

A global, functional analysis of the *Ensifer* sp. M14 proteome was performed using COG categories, and the proteome was compared with closely related species to identify general functional biases. This analysis was performed with the goal of identifying recently acquired genomic islands that may contribute to the adaptation of *Ensifer* sp. M14 to the gold mine environment. When compared with *Ensifer* sp. A49, *E. adhaerens* OV14, and *E. adhaerens* Casida A, no statistically significant biases (pairwise Fisher's exact tests, *p* > 0.05 in all cases) in COG category abundances were detected in

the *Ensifer* sp. M14 proteome (Figure 3A). However, there was a slight, but statistically insignificant (pairwise Fisher's exact tests, *p* > 0.05), enrichment in inorganic ion transport and metabolism (COG P) in the proteomes of *Ensifer* sp. M14 and *Ensifer* sp. A49 compared to the other two strains (Figure 3A). These results suggest no gross functional changes in the *Ensifer* sp. M14 genome occurred during adaptation to growth in the Zloty Stok gold mine, at least at the general level of COG categories.

**Figure 3.** General features of the genome of *Ensifer* sp. M14 and related strains. (**A**) The percentage of proteins encoded by each strain annotated with each COG (Cluster of Orthologous Genes) functional category. COG categories not represented in the proteome are excluded from the graph. COG category definitions are provided in Table S4. (**B**) A Venn diagram indicating the number of genes shared among these four strains, as extracted from the pangenome of the 46 strains shown in Figure 2. (**C**) A circular plot, prepared with Circos version 0.67-7 [62], showing the scaffolds of the *Ensifer* sp. M14 assembly (outer black curved lines) including the plasmids, and the predicted coding sequences on the positive strand (outer ring) and negative strand (inner ring). Scaffolds are drawn proportional to their size, and they are presented in the order they are numbered. Scaffold 4 (chromosome), 8 (pSinB), and 11 (pSinA) are labelled. The coding regions are colored according to their conservation level, with red indicating genes unique to *Ensifer* sp. M14, and yellow indicating species common and unique to *Ensifer* spp. M14 and A49. Some multi-gene loci unique to M14 (red asterisks) or unique to M14 and A49 (blue asterisks) are indicated.

Despite the similarity in COG abundances, the *Ensifer* sp. M14 genome contains a large number of unique genes. There are 899 genes found in *Ensifer* sp. M14 but not in *Ensifer* sp. A49, *E. adhaerens* OV14, or *E. adhaerens* Casida A, while an additional 812 are found in M14 and A49, but not OV14 or Casida A (Figure 3B). Of the 899 genes specific to *Ensifer* sp. M14, 656 (9.4% of the genome) were not detected in any of the other 45 *Sinorhizobium*/*Ensifer* strains included in the phylogenetic analysis (Data Sets S1 and S4). Five hundred and ninety of the 656 unique proteins had a blast hit (e-value ≤ 1e−10) when queried against the NCBI non-redundant protein database, consistent with the corresponding genes being real genes that were likely acquired from other organisms through horizontal gene transfer (HGT). Mapping the location of the 656 unique genes across the assembly revealed the presence of several putative genomic islands (GIs) likely acquired through recent HGT since the divergence of *Ensifer* sp. M14 from *Ensifer* sp. A49 (Figure 3C, Data Set S4). Scaffolds 11 and 8, which correspond to the pSinA and pSinB plasmids, respectively, were not surprisingly enriched in unique genes, and together account for 217 (33%) of the unique genes. As described in detail elsewhere, these plasmids carry numerous functions associated with arsenic oxidation [10] and heavy metal resistance [10,12]. Of the 439 unique genes spread among the other scaffolds, 309 (70.4%) were annotated as hypothetical genes. Little else of interest was detected among the unique genes (Data Set S4); however, scaffold 36 was predicted to encode a zinc transporting ATPase, and a few genes related to stress resistance or drug resistance were found (discussed later). Overall, these results suggest that essentially all of the recently acquired traits associated with heavy metal resistance, arsenic oxidation, and adaptation to the stressful conditions of the Zloty Stock gold mine are associated with the pSinA and pSinB plasmids.

### *3.4. Metabolism of Ensifer sp. M14*

Detailed phenotypic characterization of *Ensifer* sp. M14 was previously reported [7]. To further evaluate (in silico) the metabolic and transport potential of *Ensifer* sp. M14, a draft metabolic reconstruction was prepared encompassing 1491 genes and 1289 gene-associated reactions (Data Set S2). As expected based on the metabolism of related organisms [63], glycolysis in *Ensifer* sp. M14 is predicted to proceed through the Entner–Duodoroff pathway (Figures S3–S5). Growth simulations using Flux Balance Analysis suggested that *Ensifer* sp. M14 has a broad metabolic capacity, with a predicted ability to catabolize 72 carbon sources, including a variety of sugars, sugar alcohols, and organic acids (Table S2). This is consistent with previous work, which found that *Ensifer* sp. M14 could grow on 12 of 16 tested carbon substrates, including glucose, xylose, and lactate [7]. The following paragraphs provide a description of several metabolic capabilities that may be relevant to survival in the stressful environment of the Zloty Stok gold mine, and/or to resistance to elevated arsenic concentrations.

### 3.4.1. Phosphate Transport

The metabolic reconstruction indicated that *Ensifer* sp. M14 encodes two copies of the PstSCAB-PhoU high-affinity phosphate transporter (*BLJAPNOD\_00112* through *BLJAPNOD\_00116*; and *BLJAPNOD\_05453* through *BLJAPNOD\_05457*). Further examination of the *Ensifer* sp. M14 genome additionally revealed two copies of the PhnCDE(T) high-affinity phosphate and phosphonate transport system (*BLJAPNOD\_04783* through *BLJAPNOD\_04786*; and *BLJAPNOD\_05447* through *BLJAPNOD\_05450*). Notably, one copy of PstSCAB-PhoU and one copy of PhnCDE(T) were adjacent to the arsenic oxidation gene cluster within pSinA. This led us to explore the presence of phosphate transport systems in other arsenic-oxidizing bacteria (AOB). Using OrthoMCL [37], orthologous proteins were identified among six strains from the family *Rhizobiaceae* (Table S3): these included three AOB (*Ensifer* sp. M14, *A. tumefaciens* 5A, and *Rhizobium* sp. NT-26), as well as three related strains that are not AOB (*E. adhaerens* OV14, *N. galegae* HAMBI 540, and *A. tumefaciens* Ach5). Thirteen proteins were found to be common and specific to the three AOB, which not surprisingly included the arsenic oxidation gene cluster [10]. Notably, included within these 13 proteins were subunits of the PstSCAB-PhoU and PhnCDE(T) transporters. While all six strains encoded orthologous versions of PstSCAB-PhoU and PhnCDE(T), all three AOB encoded additional copies adjacent to their arsenic

oxidation loci. Examining the genomes of three additional diverse AOB (*H. arsenicoxydans* ULPAs1, *A. arsenitoxydans* SY8, and *P. stutzeri* TS44) revealed that the first two also contained a second copy of the PstSCAB transporter in close proximity to arsenite related genes.

Based on the above results, we predict that phosphate transport genes are commonly associated with arsenite resistance loci [64]. Arsenates and phosphate are chemical analogs, with the toxicity of arsenic being a result of arsenic replacing phosphate in key biological molecules [65]. Similarly, arsenic competes with phosphate for transport through phosphate transport systems, including the PstSCAB and PhnCDE(T) systems [66–68], potentially resulting in phosphate starvation. However, the phosphate periplasmic binding proteins of at least some PstSCAB-PhoU systems, such as from the arsenic-resistant strain *Halomonas* strain GFAJ-1, display a strong preference for binding phosphate over arsenic [68]. Thus, the presence of additional high-affinity phosphate systems in AOB may be a mechanism to increase the rate (and selectivity) of phosphate import, thereby reducing the toxic effects of elevated environmental arsenic concentrations.

### 3.4.2. Sulfur Metabolism

We evaluated sulfur metabolism by *Ensifer* sp. M14, as sulfur compounds, such as sulfide, can be abundant in gold mines, and the arsenic oxidase enzyme contains an iron-sulfur subunit [64]. *Ensifer* sp. M14 appears to have a variety of mechanisms for sulfate assimilation. Based on the metabolic reconstruction, the genome is predicted to encode multiple sulfate and thiosulfate transporters. It is further predicted to encode several putative thiosulfate sulfurtransferases and a hydrogen sulfide oxidoreductase (*BLJAPNOD\_03089*); in contrast, a sulfite oxidoreductase was not identified. Genes *BLJAPNOD\_05764* through *BLJAPNOD\_05768* may encode for the transport and metabolism of taurine, while *BLJAPNOD\_05769* may encode the TauR taurine transcriptional regulator. *Ensifer* sp. M14 is also predicted to encode an alkanesulfonate monoxygenase (*BLJAPNOD\_06609*). At least one copy of each of the subunits of the SsuABC alkanesufonate ABC-type transporter are also predicted to be encoded in the genome; however, no locus appeared to contain all three.

### 3.4.3. One-Carbon Metabolism

*Ensifer* sp. M14 is capable of growing with carbon dioxide or bicarbonate as the sole source of carbon [7], although the underlying metabolic pathway for this capability has not been examined. The metabolic reconstruction identified a putative formamide amidohydrolase (*BLJAPNOD\_04973*) and putative formate dehydrogenases (*BLJAPNOD\_00952 and BLJAPNOD\_03433*), suggestive of the utilization of these one-carbon compounds. No clear evidence for genes associated with methanol or methylamine metabolism were found. However, the mechanism underlying one-carbon metabolism remains unclear. Unlike *S. meliloti* [69], *Ensifer* sp. M14 does not appear to encode the Calvin–Benson–Bassham cycle, nor were we able to identify any of the complete carbon-fixation pathways [70]. However, multiple enzymes potentially involved in the incorporation of bicarbonate were identified. These include putative acetyl-CoA carboxylases (*BLJAPNOD\_03269*, *BLJAPNOD\_04937*, *BLJAPNOD\_04938*), a putative 3-oxopropanoate oxidoreductase (*BLJAPNOD\_03990*), putative propanoyl-CoA carboxylases (*BLJAPNOD\_06206, BLJAPNOD\_06208*), a putative pyruvate carboxylase (*BLJAPNOD\_00700*), and a phosphoenolpyruvate carboxylase (*BLJAPNOD\_01050*).

### 3.4.4. Iron Transport and Metabolism

Due to the involvement of iron in arsenic oxidation, the transport and metabolism of this metal was examined. *Ensifer* sp. M14 is predicted to encode several transporters of iron or iron containing compounds. The genes *BLJAPNOD\_01755* and *BLJAPNOD\_01831* are predicted to encode a ferrous iron (Fe2+) permease (EfeU) and a ferrous iron efflux pump (FieF), respectively. Genes *BLJAPNOD\_05889* through *BLJAPNOD\_05891* may encode a FecBDE ferric dicitrate transporter, while BLJAPNOD\_05888 may encode the FecA ferric dicitrate outer membrane receptor protein. The genes *BLJAPNOD\_00861* through *BLJAPNOD\_00863* may encode a second ferric dicitrate transporter. Additionally, the genes *BLJAPNOD\_05777*, *BLJAPNOD\_05780*, and *BLJAPNOD\_05781* may form an ABC-type transport system for iron or an iron complexes. Moreover, three putative FhuA ferrichrome (iron containing siderophore) transporting outer membrane proteins (*BLJAPNOD\_04144*, *BLJAPNOD\_04445*, *BLJAPNOD\_05778*), and a FcuA ferrichrome receptor (*BLJAPNOD\_05962*) are predicted to be encoded in the genome. A putative FepCDG ferric enterobactin transporter (*BLJAPNOD\_04147*, *BLJAPNOD\_04148*, *BLJAPNOD\_04149*) and a PfeA enterobactin receptor (*BLJAPNOD\_05560*) are also annotated. Aside from transport, *Ensifer* sp. M14 is predicted to encode a ferric reductase (*BLJAPNOD\_01976*–*fhuF*), a ferrous oxidoreductase (*BLJAPNOD\_01631*), and a ferric-chelate reductase (*BLJAPNOD\_02273*). Additionally, the five gene operon (*BLJAPNOD\_05798-BLJAPNOD\_05802*) was predicted (using antiSMASH [55]) to encode a siderophore (aerobacin-like) biosynthetic pathway. Finally, the ferric uptake regulator (Fur) is predicted to be encoded by *BLJAPNOD\_00930*.

### 3.4.5. Halotolerance

The *Ensifer* sp. M14 genome was searched for genes relevant to halotolerance as *Ensifer* sp. M14 has been shown to grow in highly saline environments with up to 20 mg L−<sup>1</sup> NaCl [10]. Examination of the *Ensifer* sp. M14 genome with antiSMASH [55] identified a 13 gene locus (*BLJAPNOD\_06859* to *BLJAPNOD\_06872*) in which 12 of the genes showed similarity to 12 of the 15 genes of a known salecan biosynthetic cluster. Salecan is a water-soluble β-glucan also produced by the salt tolerant strain *Agrobacterium* sp. ZX09 [71]. Thus, this locus in *Ensifer* sp. M14 may encode for the biosynthesis of salecan, or another carbohydrate, that contributes to halotolerance. Additionally, *Ensifer* sp. M14 is predicted to be capable of synthesizing the compatible solute betaine from choline using the BetA (*BLJAPNOD\_01468*, *BLJAPNOD\_03726*, *BLJAPNOD\_06536*) and BetB (*BLJAPNOD\_00678*, *BLJAPNOD\_03725*, *BLJAPNOD\_05671*) pathway, as well as from choline-O-sulfate with BetC (*BLJAPNOD\_02271*, *BLJAPNOD\_03724*). The genome is further predicted to encode numerous proteins related to glycine betaine and proline betaine transport. Finally, as previously reported [10], pSinA encodes a putative NhaA pH-dependent sodium/proton antiporter (*BLJAPNOD\_05431*), which may contribute to adaptation to high salinity [72].

### 3.4.6. Heavy Metal Resistance

*Ensifer* sp. M14 displays high resistance to numerous heavy metals [7]. Previous work identified eight modules related to heavy metal resistance on the pSinB replicon of *Ensifer* sp. M14 [12]. These modules were involved in resistance to arsenic, cadmium, cobalt, copper, iron, mercury, nickel, silver, and zinc [12]. Additionally, pSinA contains a locus involved in resistance to cadmium, zinc, cobalt, and mercury [10]. Our analyses reported above suggested that the majority, if not all, genes relevant to adaptation to the heavy metal-rich environment in the Zloty Stok gold mine are located on the pSinA and pSinB plasmids [10,12].

### *3.5. Biosafety Considerations of Ensifer sp. M14*

The *Sinorhizobium*/*Ensifer* group of bacteria contain numerous plant symbionts and other biotechnologically relevant strains, but it lacks known pathogens. Considering this, and the observation that none of the genomic islands detected in *Ensifer* sp. M14 appear to be pathogenicity islands, it is unlikely that *Ensifer* sp. M14 is pathogenic. Therefore, the environmental release of *Ensifer* sp. M14 is not expected to pose a biosafety risk from that perspective. Additionally, analysis of the secondary metabolism of *Ensifer* sp. M14 with antiSMASH [55] did not identify antibiotic synthesis loci. However, *Ensifer* sp. M14 may carry several antimicrobial resistance (AMR) genes. The analysis applying the RGI analyzer revealed the presence of 12 putative antibiotic resistance genes/gene clusters (Table 2). It is worth mentioning that the best hits were found for four *acrAB(-TolC)* modules encoding resistance-nodulation-cell division (RND) type multidrug efflux systems, while the remaining eight genes were much more divergent compared with the reference proteins (they were detected only when

applying the LOOSE algorithm of the RGI analyzer). This may suggest that these hits are accidental, and that the identified genes are not truly AMR genes, or that these are novel, emergent threats and more distant homologs of known reference genes.


**Table 2.** Putative antimicrobial resistance genes found in the *Ensifer* sp. M14 genome.

The most significant hits, defined with the usage of the STRICT algorithm of the RGI analyzer, are bolded. Abbreviations: AMP—ampicilin; C—chloramphenicol; CN—gentamicin; CFM—cefixime; CTX—cefotaxime, CRO—ceftriaxone; CIP—ciprofloxacin; TE—tetracycline; TM—trimethoprim; MXF—moxifloxacin; RIF—rifampicin; R—resistant; S—susceptibility; S/R—inability of interpretation of the result (threshold value).

Previous analyses revealed that the closely related organism *E. adhaerens* OV14 displays resistance to numerous antibiotics, including, among others, ampicillin, spectinomycin, kanamycin, and carbenicillin [73]. Therefore, to check whether the predicted antibiotic resistance genes truly associated with antibiotic resistance in *Ensifer* sp. M14, the MICs of 11 antibiotics were determined using Etests. Results from the Etests showed that *Ensifer* sp. M14 is resistant to ampicillin (MIC: 12.0 mg L<sup>−</sup>1), chloramphenicol (MIC: 8.0 mg L−1), and rifampicin (MIC 4.0 mg L−1), while it is susceptible to cefixime, cefotaxime, ceftriaxone, ciprofloxacin, gentamicin, moxifloxacin, and trimethoprim. In the case of tetracycline, the MIC values fluctuated around the threshold for classification as resistant (1–4 mg L−1); hence, precise interpretation of this result is not possible. Resistance to antibiotics belonging to the penams, phenicols, and rifamicyns families may be explained by the presence of efflux pumps belonging to the RND family. These multidrug resistance systems are highly prevalent in Gram-negative bacteria, and play an important role in resistance to various types of stress factors, including antibiotics [74]. It is also worth mentioning, that environmental isolates of *Alphaproteobacteria* usually possess several copies of genetic modules encoding RND type multidrug efflux systems, which may be linked with their adaptation to the heterogeneity of the soil habitat [75,76]. Therefore, we think that the environmental release of *Ensfier* sp. M14 is unlikely to pose a biosafety risk.

### *3.6. Development of a Pilot-Scale Installation for Arsenic Bioremediation*

The genomic analyses suggested that *Ensifer* sp. M14 contains several genetic features that may allow it to be successfully used in environmental bioremediation applications. In addition, previous experimental studies demonstrated that this strain can efficiently transform As(III) into As(V) (24 h of residence time was sufficient to oxidize 5 mg L−<sup>1</sup> of As(III) in the laboratory) [9]. We therefore attempted to prepare an installation for environmental bioremediation of arsenic contaminated water involving *Ensifer* sp. M14. The purification of arsenic contaminated waters constitutes a serious environmental challenge, as most of the available chemical and physical methods are dedicated to the selective removal of As(V), and are inefficient with regard to As(III). Thus, the aim of the microbiological module of the installation was to harness the arsenite oxidation capabilities of *Ensifer* sp. M14 to ensure efficient oxidation of As(III) to facilitate its subsequent removal. We reasoned that combining a biological approach with an appropriate physicochemical process (i.e., adsorption) could overcome the constraints and reservations of the conventional methods dedicated to the removal of arsenic from contaminated waters [77,78].

### *3.7. The Activity and Characterization of the Microbiological Module of the Pilot-Scale Installation*

In our preliminary study [9], we observed that efficient functioning of the laboratory-scale installation required a high density of *Ensifer* sp. M14 (OD600 between 0.1 and 0.2). This is in part because the quantity of *Ensifer* sp. M14 usually decreases quite intensively during the first hours/days of continuous culturing in the bioreactor [9]. Although appropriate growth conditions and length of residence time during continuous cultures were previously determined [9], the move from the laboratory-scale to pilot-scale installation meant it was necessary to re-evaluate them. In particular, replacement of the synthetic medium by natural arsenic contaminated water, as well as increasing the scale of application, may result in a deceleration of bacterial growth and a decrease in the efficiency of the biooxidation processes [79].

### 3.7.1. Microbial Growth and Efficiency of Arsenic Biooxidation in the Bioreactor

Using the start-up procedures described in the Materials and Methods, the initial quantity of bacteria in the bioreactor after yeast extract augmentation was about 108 CFU mL−<sup>1</sup> (Figure 4). The value was almost nine orders of magnitude higher compared to raw arsenic-contaminated water, where the CFU mL−<sup>1</sup> (when plated on Luria-Bertani agar medium) was about 100.

As expected based on our preliminary study [9], the density of bacteria decreased systematically during the first few days of operation, reaching a density on the magnitude of 10<sup>3</sup> CFU mL−<sup>1</sup> on day seven (Figure 4A). After this point, the density of bacteria largely stabilized, with the exception of a few days (days 17–20), when an ~100-fold drop in bacterial density was observed (Figure 4A). A bacterial concentration of 103 CFU mL−<sup>1</sup> in the bioreactor generally appeared sufficient for efficient biooxidation of the arsenite in the contaminated water, as there was generally little to no arsenite detected in the water following passage through the bioreactor (Figure 4D). The exceptions were five of the nine measurements taken between days 15 and 31, inclusive, when arsenite accounted for up to 62.86% of the total arsenic concentration; this corresponded with the drop in the density of bacteria within the bioreactor (Figure 4A). Thus, the low arsenite concentration throughout the majority of the experiment suggests that the microbiological module efficiently converted the As(III) to As(V).

Recently, Tardy et al. [52] showed that efficient arsenite biooxidation in environmental samples of water at 20 ◦C occurred after eight days of culture (batch experiment), and the quantity of bacteria at the end of their experiment was about 105 CFU mL−1. On the other hand, Kamde et al. [80] reported that arsenic removal was most intensive when the quantity of bacteria was about 28 CFU mL−<sup>1</sup> (batch cultures with the use on synthetic medium). The higher quantity of bacteria in the abovementioned papers in comparison with our study is presumably related to differences in culture conditions (various media and/or culturing methods).

Our data (Figure 4A,D) is also consistent with a relationship between the quantity of *Ensifer* sp. M14 and the efficiency of arsenic biooxidation, as were our preliminary experiments in batch cultures (data not shown). Indeed, many studies have observed a positive correlation between the density of bacteria and the rate of metal metabolism or biotransformation for arsenic compounds and other elements [80–82].

**Figure 4.** Parameters of the water in the bioreactor of the microbiological module. The graphs display (**A**) the quantity of bacteria, (**B**) the pH of the water, (**C**) the redox potential of the water, and (**D**) the concentration of As(III) (blue) and total arsenic (red) in the water.

### 3.7.2. Physical and Chemical Characterization of the Bioreactor

Previous studies have observed that there is a relationship between pH and redox potential with the arsenite/arsenate ratio; arsenites are the predominant form in reducing conditions and lower pH values, as the concentration of the arsenate form increases, both pH and redox potential also increase [83,84]. We therefore evaluated the pH and the redox potential in the treated water. For both parameters, the biological treatment had a small but noticeable effect. In the raw water, the pH and the redox potential were 7.48 and 170.90 mV, respectively [56]. In the case of the pH, the value in the bioreactor systematically increased up to the eighth day, with the treated water reaching a pH of 8.09 (Figure 4B). The pH returned to 7.60 by day 17, following which the pH stabilized in the range of 7.60 to 7.65 until the end of the experiment (day 40). In general, the redox potential remained relatively stable (Figure 4C). For the first three days, a value around 155.00 mV was observed, following which the redox potential increased and stabilized (with a slight, gradual decrease) within a range from 177.00 mV and 193.00 mV, with the exception of day 20. Water for human consumption is expected to have a pH in the range of 6.50–9.00 [85] and a redox potential between 100 and 300.00 mV [86]. Thus, both the pH and the redox potential of the water treated with our installation fell within the acceptable range for drinking water.

### *3.8. Effectiveness of the Adsorption Module of the Pilot-Scale Installation*

Granulated bog iron ores are characterized by high arsenic adsorption capacity (up to 5.72 mg kg−1, depending on the adsorbate concentration), short residence time (20 min) [56], they display high chemical stability, and they are resistant to bioweathering processes [87]. These properties allow this material to function as an effective adsorbent for removal of arsenics from contaminated water in both passive and active remediation systems, as demonstrated in our earlier work [56]. Here, we have coupled the use of granulated bog iron ores as an input to the adsorption module as well as the microbiological module described above, as a way to ensure the efficient conversion of As(III) to As(V) by *Ensifer* sp. M14, followed by the removal of As(V) by the bog iron ores. The pre-conversion of As(III) to As(V) is important as bog iron ores saturated with As(V) display higher chemical stability than bog iron ores saturated with As(III) [87].

Treatment of the arsenic contaminated water with the pilot-scale installation led to a dramatic decrease in arsenic concentrations, going from 2400 μg L−<sup>1</sup> in the raw water to less than 10 μg L−<sup>1</sup> (Figure 5). Analysis of the breakthrough curves for each of the adsorption columns indicated that the adsorbent in none of the columns reached equilibrium saturation during the 40-day experiment (Figure 5). Equilibrium saturation is herein defined as the maximum adsorption capacity (full saturation) of the adsorbent at a given concentration of the adsorbate; i.e., when the arsenic concentration in the input and output water is equal. Upon reaching equilibrium saturation, the adsorbent would be completely consumed and unable to further remove arsenic from the water, and it would therefore require regeneration or replacement. As the total arsenic concentration in water after each column was lower than the water entering the column, none of the columns reached equilibrium saturation. Thus, under the tested environmental conditions, the pilot-scale installation is expected to have been able to effectively continue the bioremediation process for much longer than the 40 days of the experiment (during which, 8 m3 of water flowed through the system).

**Figure 5.** Arsenic adsorption breakthrough curves. The arsenic adsorption breakthrough curves for each column of the adsorption module are shown. Total arsenic concentrations in the raw water (black), and after column I (blue), column II (red), and column III (purple) are shown.

In Poland, the Regulation of the Polish Ministry of the Environment [88] currently sets the upper limit for arsenic contamination in water for use in technological purposes at 100 μg L−1. In the experiment reported here, the arsenic concentration in the treated water remained below 100 μg L−<sup>1</sup> for the first ten days of the experiment (Figure 5), and never exceeded 220 μg L−<sup>1</sup> during the 40 day test. Thus, at least the first 2.0 m<sup>3</sup> of water treated by pilot-scale installation was below the Polish limit for use in technological purposes. However, if pooling the treated water (and thus averaging the

arsenic concentration), it is likely that the cumulative concentration of arsenic in the 4.0 m3 of water treated over the first 20 days remained below the limit.

The local adsorption capacity of the adsorbent varied between the columns and depended on the arsenic concentration of the inflowing water. The adsorbent from the first column was characterized by the highest adsorption capacity, which was 0.500 mg kg−1. Adsorbent placed in the second and third columns had lower adsorption capacities of 0.031 and 0.021 mg kg−1, respectively. Likely, these differences are due to the later columns adsorbing less arsenic and being farther from reaching equilibrium saturation. The adsorption capacities recorded in the current study were significantly lower than those described in our previous work, presumably due to the adsorbent not reaching equilibrium saturation [56].

### **4. Conclusions**

Here, we reported the draft genome sequence of *Ensifer* sp. M14 in order to gain insights into the genomic adaptation of this organism to the stressful environment of the abandoned Zloty Stok gold mine from which it was isolated. In addition, we were interested in the genetic basis of the strains arsenic oxidation and resistance capabilities, resistance to arsenic and other heavy metals, and the biosafety of the strain for use in biotechnological applications. The results revealed hundreds of genes present in *Ensifer* sp. M14 that are not found in related species, and these genes are often colocalized in genomic islands. However, the majority of these genes encoded hypothetical proteins of unknown function. Based on the genome sequence, it appears that the majority of the genes have been acquired to deal with the hostile environment of the Zloty Stok gold mine, i.e., conferring resistance to heavy metals, and enabling arsenic oxidation, are located on the self-transmissible pSinA and pSinB megaplasmids. Additionally, analysis of the *Ensifer* sp. M14 genome suggested that this strain should be safe for use in biotechnology and bioremediation. However, it was noted that several putative antibiotic resistance genes are present in the genome, as is also true for the related strain *Ensifer adhaerens* OV14 that is used in biotechnological applications [89]. This property of *Ensifer* sp. M14 should be kept in mind during its application in order to limit the spread of antimicrobial resistance. The results of these genomic analyses provide hints into the genetic potential of *Ensifer* sp. M14. They will help focus future experimental research aimed at further characterizing the biology of this organism, and may contribute to the development of procedures for large-scale cultivation of this strain.

This study also reports the construction and validation of a pilot-scale installation designed for the remediation of arsenic contaminated waters. This novel installation couples a microbiological module, based on the arsenic oxidation abilities of *Ensifer* sp. M14, with an adsorption module, based on the use of granulated bog iron ores. The underlying principle is to use *Ensifer* sp. M14 to efficiently oxidize the As(III) ions to As(V), followed by the removal of the As(V) through adsorption by the bog iron ores. Characterization of the arsenic contaminated water following passage through the microbiological module generally revealed little to no As(III), consistent with the *Ensifer* sp. M14 generally ensuring effective conversion of As(III) to As(V). Additionally, a dramatic decrease (from 10-fold to greater than 250-fold) in the arsenic concentration of the water was observed following passage of this water through the adsorption module. These results therefore confirm the effectiveness of the tested installation for the remediation of arsenic contaminated waters, which pose risks to both the environmental and human health. Future work will be aimed at further developing and optimizing this system, which could involve, for example, the addition of beads to the reactor containing *Ensifer* sp. M14 biofilms.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/9/8/379/s1. Table S1: Accession numbers for all genomes used in this work, Table S2: In silico test of the metabolic capacity of *Ensifer* sp. M14, Table S3: Orthologous groupings of six *Rhizobiaceae* strains, including three AOB and three strains that are not AOB, Table S4: COG category descriptions, Figure S1: Average nucleotide identity matrix. A matrix of the two-way ANI values for 46 *Sinorhizobium*/*Ensifer* strains is shown. Clustering was performed along both axes using hierarchical clustering with Pearson distance and average linkage, Figure S2: Average amino acid identity matrix. A matrix of the two-way AAI values for 46 *Sinorhizobium*/*Ensifer* strains is shown. Clustering was performed along both axes using hierarchical clustering with Pearson distance and average linkage, Figure S3: Entner–Duodoroff pathway and the pentose phosphate pathway. A modified version of the KEGG pathway map ko00030 [90] displaying the Entner–Duodoroff pathway and the pentose phosphate pathway is shown. Reactions encoded by the *Ensifer* sp. M14 genome are colored green; those in white are missing. The figure was prepared using the KAAS webserver [91] using BLAST search with the bi-directional best hit assignment method, and with the default organism list for 'prokaryotes' plus *Sinorhizobium meliloti* Rm1021, Figure S4: Gluconeogenesis. A modified version of the KEGG pathway map ko00010 [90] displaying the pathway for gluconeogenesis is shown. Reactions encoded by the *Ensifer* sp. M14 genome are colored green; those in white are missing. The figure was prepared using the KAAS webserver [91] using BLAST search with the bi-directional best hit assignment method, and with the default organism list for 'prokaryotes' plus *Sinorhizobium meliloti* Rm1021, Figure S5: Tricarboxylic acid cycle. A modified version of the KEGG pathway map ko00020 [90] displaying the tricarboxylic acid (TCA) cycle is shown. Reactions encoded by the *Ensifer* sp. M14 genome are colored green; those in white are missing. The figure was prepared using the KAAS webserver [91] using BLAST search with the bi-directional best hit assignment method, and with the default organism list for 'prokaryotes' plus *Sinorhizobium meliloti* Rm1021, Data Set S1: *Sinorhizobium*/*Ensifer* gene presence and absence. This file contains the gene presence/absence output data from Roary for the pangenome analysis of 46 *Sinorhizobium*/*Ensifer* strains. Details on the information provided in the file is available at: https://sanger-pathogens.github.io/Roary/, Data Set S2: Metabolic reconstruction of *Ensifer* sp. M14. This archive contains the expanded, draft metabolic reconstruction of *Ensifer* sp. M14. The reconstruction is provided in COBRA format as a MATLAB file, as well as in a table within an Excel workbook. A readme file is included to explain the contents, Data Set S3: PhiSpy phage prediction. This file contains the PhiSpy phage prediction output for all *Ensifer* sp. M14 genes, as well as separate sheets for each of the putative prophage loci. Details on the information provided in the file is available at: https://github.com/linsalrob/PhiSpy, Data Set S4: Functional annotation of the *Ensifer* sp. M14 genome. This file contains the genome annotation and the eggNOG-mapper output for three sets of genes: (i) all genes in the *Ensifer* sp. M14 genome; (ii) all genes unique to the *Ensifer* sp. M14 genome; and (iii) all genes unique and common to the *Ensifer* sp. M14 and A49 genomes. Details on the eggNOG-mapper output provided in the file is available at: https://github.com/jhcepas/eggnog-mapper/wiki/Results-Interpretation.

**Author Contributions:** Conceptualization, G.C.d., K.D., A.M., L.D., and L.Dr.; Methodology, G.C.d., K.D., and J.K.; Software, G.C.d., A.M., C.F., and L.Dz.; Validation, G.C.d., K.D., A.M., L.Dz., and L.Dr.; Formal Analysis, G.C.d. and K.D.; Investigation, K.D., W.U., A.G., G.R., and T.B.; Resources, K.D., A.M., and L.Dr.; Data Curation, G.C.d., A.M., C.F., and L.Dz.; Writing-Original Draft Preparation, G.C.d., K.D., and L.Dr., Writing-Review & Editing, W.U., A.M., L.Dz., T.B., and G.R.; Visualization, G.C.d., K.D., and J.K.; Supervision, A.M. and L.Dr.; Project Administration, A.M. and L.Dr.; Funding Acquisition, K.D., A.M., and L.Dr.

**Funding:** This research was funded by National Centre for Research and Development (Poland) grant number LIDER/043/403/L-4/12/NCBR/2013, National Science Center (Poland) grant no. 2016/23/N/NZ9/01655 and intramural funding from the University of Florence, call "PROGETTI STRATEGICI DI ATENEO ANNO 2014".

**Acknowledgments:** G.C.d. was supported by a Post-Doctoral Fellowship from the Natural Sciences and Engineering Council of Canada. K.D. was supported by the European Molecular Biology Organization in the frame of the EMBO Short-Term Fellowship program [grant number 7376].

**Conflicts of Interest:** The authors declare no conflicts of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article*

## **Possible Role of Envelope Components in the Extreme Copper Resistance of the Biomining** *Acidithiobacillus ferrooxidans*

### **Nia Oetiker 1, Rodrigo Norambuena 1, Cristóbal Martínez-Bussenius 1, Claudio A. Navarro 1, Fernando Amaya 2, Sergio A. Álvarez 2, Alberto Paradela <sup>3</sup> and Carlos A. Jerez 1,\***


Received: 30 May 2018; Accepted: 3 July 2018; Published: 10 July 2018

**Abstract:** *Acidithiobacillus ferrooxidans* resists extremely high concentrations of copper. Strain ATCC 53993 is much more resistant to the metal compared with strain ATCC 23270, possibly due to the presence of a genomic island in the former one. The global response of strain ATCC 53993 to copper was analyzed using iTRAQ (isobaric tag for relative and absolute quantitation) quantitative proteomics. Sixty-seven proteins changed their levels of synthesis in the presence of the metal. On addition of CusCBA efflux system proteins, increased levels of other envelope proteins, such as a putative periplasmic glucan biosynthesis protein (MdoG) involved in the osmoregulated synthesis of glucans and a putative antigen O polymerase (Wzy), were seen in the presence of copper. The expression of *A. ferrooxidans mdoG* or *wzy* genes in a copper sensitive *Escherichia coli* conferred it a higher metal resistance, suggesting the possible role of these components in copper resistance of *A. ferrooxidans*. Transcriptional levels of genes *wzy*, *rfaE* and *wzz* also increased in strain ATCC 23270 grown in the presence of copper, but not in strain ATCC 53993. Additionally, in the absence of this metal, lipopolysaccharide (LPS) amounts were 3-fold higher in *A. ferrooxidans* ATCC 53993 compared with strain 23270. Nevertheless, both strains grown in the presence of copper contained similar LPS quantities, suggesting that strain 23270 synthesizes higher amounts of LPS to resist the metal. On the other hand, several porins diminished their levels in the presence of copper. The data presented here point to an essential role for several envelope components in the extreme copper resistance by this industrially important acidophilic bacterium.

**Keywords:** *Acidithiobacillus ferrooxidans*; copper resistance; biomining; envelope components; proteomics; lipopolysaccharide

### **1. Introduction**

*Acidithiobacillus ferrooxidans* is a gram-negative, acidophilic, chemolithoautotrophic bacterium able to use ferrous iron, reduced species of sulfur or metal sulfides as energy sources [1–5]. These bacteria are able to grow at high concentrations of several metals. This is an important property since they are used in biomining processes where copper concentrations are in the range of 15 to 100 mM [6–8]. Furthermore, these microorganisms can be used to exploit these natural resources sustainably [9]. Current knowledge indicates that *A. ferrooxidans* uses key elements involved in copper resistance in all bacteria [10–12], but in addition, it may have a broader repertoire of these known copper resistance determinants [13,14].

In the biomining environment, copper and other toxic metals are present in concentrations that are one to two orders of magnitude greater than those tolerated by neutrophils [6,15–18]. Most likely, the microorganisms forming part of the biomining consortium have developed additional strategies to resist the harsh conditions in which they live, and their study is therefore of great interest [19,20].

*A. ferrooxidans* ATCC 53993 is much more resistant to copper and other metals than *A. ferrooxidans* ATCC 23270. Both strains have the same copper resistance determinants but strain ATCC 53993 contains a genomic island (GI) having 160 extra genes, some of which code for additional copies of proteins involved in copper tolerance [13,14].

The response to copper of both *A. ferrooxidans* ATCC 53993 and ATCC 23270 was previously compared at 40 mM CuSO4 [14,21,22]. These preliminary studies were done by using ICPL (isotope-coded protein labeling) quantitative proteomics [22] and showed that strain ATCC 23270 synthesized much more oxidative-stress-related proteins than strain 53993 in response to copper, clearly indicating that the former strain is much more sensitive to the metal [22]. A high overexpression of RND (Resistance-Nodulation-Division) efflux systems and copper periplasmic chaperones CusF were seen in both strains subjected to copper. However, in strain ATCC 53993 both of its additional genes present in its genomic island were also overexpressed. This behavior suggested a possible explanation for the much higher copper resistance of strain ATCC 53993. In addition, changes in the levels of the respiratory system copper-binding proteins AcoP, Rus and several other proteins with predicted functions suggested that numerous metabolic changes are involved in controlling the effects of the toxic metal in strain ATCC 53993 [22].

To understand in more detail the reason by which *A. ferrooxidans* ATCC 53993 stands higher copper concentrations compared with strain 23270, iTRAQ (isobaric tag for relative and absolute quantitation) proteomics, transcriptional expression of genes of interest and functional assays were used in the current report. Increased levels of novel possible copper resistance determinants present in the envelope of *A. ferrooxidans* ATCC 53993 such as outer membrane proteins, the periplasmic glucans synthesizing protein MdoG and proteins involved in lipopolysaccharide (LPS) synthesis, amongst others, were found in cells grown in the presence of copper. In addition, determination of the relative amounts of LPS present in the cells of each *A. ferrooxidans* strain also supports the idea that these polymers may also have an important role in copper resistance in these biomining bacteria.

### **2. Materials and Methods**

### *2.1. Bacterial Strains and Growth Conditions*

*A. ferrooxidans* strains ATCC 53993 and ATCC 23270 were grown at 30 ◦C in liquid 9 K medium containing ferrous sulfate (33.33 g/L) with an initial pH of 1.45 as previously described [23] and in absence or presence of CuSO4. Copper concentrations between 40 and 300 mM were used depending on the experiment. In some experiments, concentrations of 100 or 200 mM were used for *A. ferooxidans* ATCC 53993 without prior adaptation since under these two conditions, similar cells numbers to control cells (in absence of copper) were obtained at their respective late exponential growth phases. At 300 mM copper strain ATCC 53993 required previous adaptation. On the other hand, to compare the effect of the metal at the same copper concentration in both strains, it was necessary to adapt strain ATCC 23270 to grow at 100 mM copper. For LPS determinations, strain ATCC 23270 was adapted to grow at 100 mM and strain ATCC 53993 to 200 mM copper. These adaptations were done starting with cells grown at 50 mM copper by increasing 5 mM copper in each successive culture until the desired concentrations were reached. After cells attained late exponential growth phases they were collected and triplicate separate cultures were employed for all experiments. Bacterial growth was determined

by measuring the increase in cell numbers by using an Olympus BX50 optical microscope (Olympus, Tokyo, Japan) and a Petroff–Hausser counting chamber (Horsham, PA, USA).

### *2.2. Preparation of Total Protein Extracts for iTRAQ Analysis*

*A. ferrooxidans* ATCC 53993 was grown with ferrous iron as oxidizable substrate until late exponential phase in absence or presence of 100 or 200 mM CuSO4. Cells were harvested by centrifugation (4000× *g* for 15 min) and washed three times by centrifugation at 4 ◦C with dilute sulfuric acid (pH 1.5). This was followed by three washes with 50 mM sodium citrate, pH 7.0 by centrifugation at 4 ◦C to remove any minor ferrous iron remaining and at the same time, to neutralize the pH before cell rupture by sonic oscillation. Cells were then resuspended in sonication buffer (50 mM Tris-HCl pH 8.0, 1 mM ethylenediaminetetra-acetic acid (EDTA) containing phenylmethylsulfonyl fluoride (PMSF) as protease inhibitor (100 μg/mL) and were disrupted by sonic oscillation during 25 min on ice by using successive 5 s pulses and pauses. Finally, the lysate was centrifuged at 10,000× *g* for 10 min to remove unbroken cells and cell debris and the total protein amount in the cell-free extract was determined [21].

### *2.3. Protein Digestion and Tagging with iTRAQ-8-Plex® Reagent*

Total protein concentration was determined using microBCA protein assay kit (Pierce, Appleton, WI, USA). For digestion, 50 μg of protein from each condition was precipitated by the methanol/chloroform method. Protein pellets were resuspended and denatured in 20 μL of 7 M urea, 2 M thiourea, 100 mM TEAB (triethylammonium bicarbonate), reduced with 2 μL of 50 mM Tris 2-carboxyethyl phosphine (TCEP) (AB SCIEX, Foster City, CA, USA), pH 8.0, at 37 ◦C for 60 min and followed by 2 μL of 200 mM cysteine-blocking reagent methyl methanethiosulfonate (MMTS) (Pierce) for 10 min at room temperature. Samples were diluted up to 120 μL with 50 mM TEAB to reduce the concentration of urea. Two μg of sequence grade-modified trypsin (Sigma-Aldrich, St. Louis, MO, USA) was added to each sample (ratio 1:25 enzyme:sample, which were then incubated at 37 ◦C overnight on a shaker. After digestion, samples were dried in a SpeedVac (Thermo Scientific, Waltham, MA, USA).

Each sample was reconstituted with 180 μL of 70% ethanol/50 mM TEAB, the different versions of the iTRAQ reagent 8-plex (AB SCIEX) were added in additional 20 μL and the mixture was incubated for 2 h at room temperature, according to the following labeling scheme: iTRAQ 113/117 reagent: control 1 and control 2 *A. ferrooxidans*; iTRAQ 115/119 reagent: *A. ferrooxidans* grown in 100 mM CuSO4, 1 and grown in 100 mM CuSO4, 2; iTRAQ 116/121 reagent: *A. ferrooxidans* grown in 200 mM CuSO4, 1 and grown in 200 mM CuSO4, 2. Two biological replicas were used in each case. After labeling, samples were combined and the reaction was stopped by evaporation in the SpeedVac.

### *2.4. Liquid Chromatography and Mass Spectrometry Analysis*

A 2-μg aliquot of the combined sample was subjected to 2D-nano Liquid Chromatography-Electrospray Ionization Tandem Mass Spectrometry LC ESI-MSMS analysis using a nano liquid chromatography system nanoLC Ultra 1D plus, (Eksigent Technologies, AB SCIEX) coupled to a Quadrupole time of flight (QTOF) type, high speed Triple TOF 5600 mass spectrometer (AB SCIEX) equipped with a nanospray source. Injection volume was 5 μL and three independent technical replicas were analyzed. The analytical column used was a silica-based reversed phase Acquity UPLC Peptide BEH C18 column 75 μm × 15 cm, 1.7 μm particle size and 130 Å pore size (Waters, Dublin, Ireland). The trap column was a C18 Acclaim PepMap (Eksigent Technologies, AB SCIEX), 100 μm × 2 cm, 5 μm particle diameter, 100 Å pore size, switched on-line with the analytical column. The loading pump delivered a solution of 0.1% formic acid in water at 2 μL/min. The nano-pump provided a flow-rate of 300 nL/min and was operated under gradient elution conditions, using 0.1% formic acid in water as mobile phase A, and 0.1% formic acid in acetonitrile as mobile phase B. Gradient elution was performed according to the following scheme: Isocratic

conditions of 96% A: 4% B for 5 min, a linear increase to 40% B in 205 min, then a linear increase to 90% B for 15 additional minutes, isocratic conditions of 90% B for 10 min and return to initial conditions in 2 min. Total gradient length was 250 min.

Data acquisition was performed with a TripleTOF 5600 System (AB SCIEX). Ionization occurred under the following conditions: Ionspray voltage floating (ISVF) 2800 V, curtain gas (CUR) 20, interface heater temperature (IHT) 150, ion source gas 1 (GS1) 20, declustering potential (DP) 85 V. All data was acquired using information-dependent acquisition (IDA) mode with Analyst TF 1.5 software (AB SCIEX). For IDA parameters, 0.25 s MS survey scan in the mass range of 350–1250 Da were followed by 25 MS/MS scans of 150 ms in the mass range of 100–1500 (total cycle time: 4 s). Switching criteria were set to ions greater than mass to charge ratio (*m*/*z*) 350 and smaller than *m*/*z* 1250 with charge state of 2–5 and an abundance threshold of more than 90 counts (cps). Former target ions were excluded for 20 s. IDA rolling collision energy (CE) parameters script was used for automatically controlling the CE.

### *2.5. Data Analysis and Statistics*

MS/MS spectra were exported to Mascot generic format (mgf) using Peak View v1.2.0.3 and searched using OMSSA 2.1.9, X!TANDEM 2013.02.01.1, Myrimatch 2.2.140 and MS-GF+ (Beta v10072) [24] against a composite target/decoy database built from the 2748 *A. ferrooxidans* sequences at UniprotKB (June 2014). Search engines were configured to match potential peptide candidates with mass error tolerance of 25 ppm and fragment ion tolerance of 0.02 Da, allowing for up to two missed tryptic cleavage sites and a maximum isotope error (13C) of 1, considering fixed MMTS modification of cysteine and variable oxidation of methionine, pyroglutamic acid from glutamine or glutamic acid at the peptide N-terminus, and modification of lysine and peptide N-terminus with iTRAQ 8-plex reagents. Score distribution models were used to compute peptide-spectrum match *p*-values [24], and spectra recovered by a false discovery rate (FDR) ≤ 0.01 (peptide-level) filter were selected for quantitative analysis. Approximately 5% of the signals with the lowest quality were removed prior to further analysis. Differential regulation was measured using linear models [25], and statistical significance was measured using *q*-values (FDR). All analyses were conducted using software from Proteobotics (Madrid, Spain) [24].

### *2.6. Extraction of Total RNA from Acidithiobacillus ferrooxidans and Complementary DNA Synthesis*

To determine the effect of copper on the expression of some genes of interest, *A. ferrooxidans* ATCC 23270 and ATCC 53993 cells were grown in absence or presence of CuSO4 until cells reached late exponential phase of growth. At this time, total RNA was extracted from each culture condition by lysing the cells as previously reported [26], except that TRIzol (Invitrogen, Carlsbad, CA, USA) was used for the extraction [27,28]. Between three to five biological replicas were used for each experimental condition. Any remaining DNA was eliminated from RNA preparations by addition of 4 U of TURBO DNA-free DNase (Ambion, Thermo Scientific) following manufacturer's instructions. For complementary DNA (cDNA) synthesis, 0.8 μg of total RNA was reverse transcribed for 1 h at 42 ◦C using ImProm-II (Promega, Madison, WI, USA) reverse transcription system, 0.5 μg of random hexamers (Promega) and 3 mM MgCl2 [28].

### *2.7. Primer Design, Real-Time PCR and Cloning of A. ferrooxidans Genes*

Primers for quantitative real time PCR (qRT-PCR) were designed using the Primer3 software [29]. After separating PCR products by electrophoresis in a 1% agarose gel (0.5× Tris–acetate–EDTA pH 8.0 buffer), no cross-amplification or non-specific bands were detected. Copper-resistance related gene expression was analyzed by qRT-PCR with either the Corbett Rotor Gene 6000 system as described previously [21] or with the 96-well PikoReal Real-Time PCR System and Thermo Scientific PikoReal Software 2.2. Efficiency of each primer pair was calculated from the average slope of a linear regression curve, which resulted from qPCRs using a 10-fold dilution series (10 pg–10 ng) of *A. ferrooxidans* DNA as template. Efficiencies between 90 and 110% were used. Quantification cycle (Cq) values were

automatically determined by Real-Time Rotor-gene 6000 PCR software (Corbett Life Sciences, Thermo Scientific/Qiagen, Hilden, Germany) or by Thermo Scientific PikoReal Software 2.2.

For transcriptional analysis of the different genes studied, a relative quantification method was used which is based in the ratio between the transcripts of a study sample (in presence of copper) versus a control sample (no copper) [30]. 16S rRNA*Af* was selected as a reference gene since its expression was found to be the most stable under our experimental conditions. To carry out the real-time PCR, 0.5 μL of 1:20 diluted cDNA or 0.5 μL of 1:200 diluted 16S rRNA*Af*, 0.2 μL of each primer (10 μM) and 5.0 μL of master mix Rotor-Gene SYBR Green PCR (Qiagen) in a final volume of 10 μL, completed with RNA-free water were used. The program used was 10 min at 95 ◦C followed by 40 cycles of 5 s at 95 ◦C and 20 s at 60 ◦C.

### *2.8. Cloning A. ferrooxidans Genes in an Expression Vector*

The functionality of different putative copper resistance genes from *A. ferrooxidans* was tested by using heterologous expression in *Escherichia coli*. A copper-sensitive *E. coli* K-12 (Δ*copA/*Δ*cusCFBA/*Δ*cueO*) mutant was transformed with vector pTrc-His2A (Invitrogen) containing the genes of interest under the control of a promoter induced by IPTG, and minimal inhibitory concentration (MIC) values of these transformants were determined as described before [14,28].

### *2.9. Lipopolysaccharide Extraction*

*A. ferrooxidans* cells grown in absence or presence of CuSO4 were harvested by centrifugation (10,000× *g* for 5 min, at 4 ◦C). Cell pellets were washed twice with sulfuric acid solution (pH 1.5) and twice with 10 mM sodium citrate (pH 7) by resuspension followed by centrifugation (9200× *g* for 1 min). Cells were then resuspended in sulfuric acid solution. To normalize the number of cells, optical density of cell suspensions was measured at 600 nm, adjusting them to an optical density of 2 (OD600nm = 2) in 1 mL of sulfuric acid solution. Cell suspensions were then centrifuged at 10,000× *g* for 5 min. A partially modified Hitchcock & Brown method for LPS extraction was used [31]. The cell pellet was resuspended in 90 μL of lysis buffer solution (2% Sodium dodecyl sulfate (SDS); 4% 2-ME; 0.5 M Tris-HCl, pH 9.0). The suspension was heated for 30 min at 100 ◦C. Lysed cells were then digested with 100 μg/mL of DNase I (Ambion) for 90 min at 37 ◦C. Samples were thereafter treated with 1 mg/mL of Proteinase K (Sigma-Aldrich) for 90 min at 60 ◦C. Finally, samples were dialyzed for 30 min against nano-pure water using a nitrocellulose membrane, 0.025 μm pore size. Dialyzed samples were finally stored at 4 ◦C for further analysis.

### *2.10. Lipopolysaccharide Quantification*

Extracted LPS was quantified by purpald assay [32]. Unsubstituted terminal vicinal glycol (UTVG) groups of the sugar residues such as Kdo and heptose in LPS can be subjected to periodate oxidation, yielding quantitative formaldehyde measurable by the purpald reagent. This assay provides the molarity of the UTVG present in LPS. LPS molarity can be found by dividing the molarity of the UTVG by the theoretical number of UTVG per LPS molecule. The numbers of UTVG present in LPS of *A. ferrooxidans* is unknown. Therefore, LPS concentration was expressed in relation to the molarity of UTVG present in each sample. The experimental procedure was carried out as previously described [33].

### **3. Results and Discussion**

### *3.1. Proteomic Analysis of the Copper Response of A. ferrooxidans ATCC 53993*

Proteins of cells grown in ferrous iron and in presence of 100 or 200 mM CuSO4 were analyzed by quantitative iTRAQ proteomics. In cells subjected to 100 mM of copper 1656 proteins were identified, of which 28 changed their levels compared to control cells grown in absence of copper. Of these proteins, 11 had higher levels and 17 showed lower amounts (Table S1). On the other hand, in cells

grown in 200 mM copper 1567 proteins were identified and 59 of these showed changes in their levels compared to the control. Seventeen showed higher levels than the control and 42 lower amounts (Table S1). This corresponds to about 2-fold more proteins changing at 200 mM than at 100 mM copper (Table S1). Most of the proteins changing at 100 mM copper were also seen to vary at 200 mM of the metal (Tables S2 and S3). The functional categories of all proteins changing in *A. ferrooxidans* 53993 are shown in Table S1 and Figure S1 and the data obtained is seen in Tables S2 and S3. Although at 200 mM copper ATCC 53993 cells grew reaching similar numbers to the control, they were apparently more affected than cells subjected to 100 mM copper since a greater number of proteins related to metabolism and protein biosynthesis decreased their levels whereas others related to energy production and copper resistance increased their amounts. Nevertheless, at 200 mM copper cells are still actively expressing the proteins related to the RND efflux systems (Table 1), as seen before at 40 mM copper sulfate [22]. In addition, an interesting group of proteins that also form part of the cell envelope changed their synthesis levels in presence of the metal (Table 1). Most of these proteins may be new possible copper resistance determinants present in *A. ferrooxidans* ATCC 53993.



*3.2. Overexpression of Resistance-Nodulation-Division Efflux Transporters and Possible Generation of Excess Acidity*

Transcriptional levels of some genes coding for possible components of the RND family of efflux transporters [10] are shown in Figure 1. These correspond to most of the genes coding for Cus system components in both *A. ferrooxidans* ATCC 53993 [20,22] and ATCC 23270 [28]. All these genes showed increased transcriptional levels in cells grown at the indicated copper concentrations. These Cus transporter systems are widely present in bacteria to remove copper from the cell [10]. *A*. *ferrooxidans* lives at an acid external pH (1–3) and its cytoplasmic pH is up to 5 units higher than external pH. This generates an elevated pH gradient across the cytoplasmic membrane that contributes to the proton motive force (PMF) comprising membrane potential (ΔΨ) and transmembrane pH difference

(ΔpH) [22,34]. RND type transporters are antiporters taking advantage of the proton gradient to efflux copper with the concomitant protons entrance to the cytoplasm. Due to its economy from the energetic point of view, these systems would be used preferentially by acidophilic microorganisms to remove intracellular copper.

**Figure 1.** Transcriptional levels of several genes coding for Cus system components in *Acidithiobacillus ferrooxidans* ATCC 53993. The transcriptional levels of genes (**A**) *cusF2*; (**B**) *cusF3*; (**C**) *cusA3*; (**D**) *cusA1*; (**E**) *cusC1* and (**F**) *cusC2* were determined at the indicated copper concentrations as described in Material and Methods section. Error bars indicate the standard deviations based on three different experimental values. Application of *t-*Student test were: \*\*\* *p* ≤ 0.001 \*\* *p* ≤ 0.01 and \* *p* ≤ 0.05.

A possible cytoplasmic acidification would be expected if these efflux pumps were excessively used by *A. ferrooxidans* in presence of high metal concentrations. Conversely, as previously pointed out [22], this acidification could be diminished by the energetic metabolism of the bacterium, since oxidation of Fe(II) by molecular oxygen as the final electron acceptor consumes protons [35]. Still, RND systems may introduce an excess of protons from the acid culture medium to the cell during copper detoxification. This idea has not been demonstrated in *A. ferrooxidans*. However, a possible increase in the extracellular pH of the growth medium could be expected during growth in the presence of copper. Figure 2A shows growth curves of *A. ferrooxidans* ATCC 53993 and how cell growth was affected at the indicated copper concentrations. Initially, there was a strong partial inhibition of growth only at 300 mM copper.

Figure 2B clearly shows an increase in external pH of the growth medium (reaching around 0.6 pH units at 300 mM copper). On the contrary, pH values of control media containing different copper concentrations and no cells inoculated showed only very minor pH variations. Clearly, whether this interesting preliminary observation is due to intracellular acidification remains to be demonstrated.

**Figure 2.** Growth medium pH changes of *A. ferrooxidans* ATCC 53993 grown in absence or presence of copper. (**A**) Cells were grown in ferrous iron medium at the indicated copper concentrations. Once cells reached stationary phase of growth (indicated by the arrow), aliquots of the cultures were taken and centrifuged to remove cells; (**B**) pH values of the media supernatants were determined and compared with pH changes of the medium containing the same copper concentrations but in absence of inoculated cells. Error bars indicate standard deviations based on three different experimental values. Application of *t*-Student test were: \*\*\* *p* ≤ 0.001 \*\* *p* ≤ 0.01 and \* *p* ≤ 0.05.

### *3.3. Changes of Several Additional Envelope Components Occur in Presence of High Copper Concentrations*

Another protein with increased levels in cells grown in presence of copper was a putative periplasmic glucan biosynthesis protein MdoG coded by *Lferr\_1075* (Table 1). This protein is involved in the synthesis of ramifications present in the osmoregulated periplasmic glucans (OPGs) in bacteria. These OPGs are present in all known proteobacteria and are formed by 5–24 D-glucose molecules bound by means of β-glycosidic bonds. The concentration of these glucans has been reported to change with variations in periplasmic osmolarity [36]. Due to their big size, OPGs are trapped in the periplasm, being unable to diffuse to the outside of cells. In *E. coli*, the carbon skeleton is synthesized by proteins coded by genes *opgG* (*mdog* orthologous) and *opgH*. OpgH is a glucosyl transferase that synthesizes the lineal skeleton of glucose units by means of β-1,2 bonds. In *E. coli*, MdoG is a 56 kDa periplasmic protein necessary for the polymerization of sugar molecules, although its function has not being completely established [36]. In *A. ferrooxidans*, putative MdoG (57.4 kDa) has been previously identified as a component of its periplasm [37]. In *E. coli* both *opgG* and *opgH* genes form part of the same operon. By analyzing the genomic context of *A. ferrooxidans* instead, it was found that *mdoG* and *mdoH* are separated by an open reading frame (ORF) coding for a protein of unknown function. On the other hand, protein MdoH did not change its levels in the results obtained here. The system for OPGs synthesis in *E. coli* involves four additional proteins (OpgD, OpgB, OpgC and OpgE) whose equivalent genes are absent in *A. ferrooxidans* genome. Nevertheless, only OpgG and OpgH are strictly necessary for the OPGs synthesis in *E. coli* [38]. OPGs biosynthesis starts with glucose transport to form glucose-6P, which is used to generate (uridine diphosphate glucose) UDP-glucose for production of OPGs via OpgH/OpgG [36]. It is known this molecule is formed by a glucose 1-phosphate uridil transferase that catalyzes the UTP and a proton addition to D-glucose-1-phosphate to generate UDP-D-glucose [39]. It can be suggested that generation of OPGs in *A. ferrooxidans* would involve also a higher UDP-glucose synthesis. Since this process consumes protons, it should alleviate excessive entrance of these cations when RND efflux pumps are heavily used to remove copper.

The CusA proton/Cu antiporter system is overexpressed in *A. ferrooxidans* subjected to copper as already seen in Figure 1, and under those conditions, a higher number of protons would be expected to enter the cytoplasm from the growth medium, as already suggested by the results shown in Figure 2. Thus, a higher synthesis of OPGs would also consume protons, in favor of keeping the normal cytoplasmic pH. Furthermore, it has been shown that an OPGs preparation acts as a blocker and a regulator of an OMPC-like porin channel selective of cations in *E. coli* [40]. On the other hand, cells unable to form OPGs showed an increased synthesis of OmpC [41]. It has also been documented that porins mediate copper entrance in *Mycobacterium tuberculosis* [42]. The existence of a relationship

between both copper entrance and porins closing or decreasing their levels of synthesis is possible then, as seen here for OprB in *A. ferrooxidans* ATCC 53993 (Table 1). Examples of this behavior were previously reported for the major *A. ferrooxidans* porin Omp40 (*Afe\_2741*) and OmpA (*Afe\_2685*) in *A. ferrooxidans* ATCC 23270 [21].

To support proteomic results, transcriptional levels of genes coding for proteins MdoG and porins, were also determined in cells grown at different copper concentrations as shown in Figure 3. The results clearly indicate increasing levels of synthesis of mRNA coding for MdoG and decreasing levels of messenger RNAs for porin genes *omp40*, *oprB* and *ompA*, confirming the proteomic results already discussed. It is, therefore, possible that lower levels of porins, together with higher OPGs amounts, constitute a defense response to extreme copper conditions as seen here in *A. ferrooxidans* ATCC 53993, an idea that should be proven.

**Figure 3.** Transcriptional levels of selected envelope genes. (**A**) *mdoG*; (**B**) *oprB*, (**C**) *ompA* and (**D**) *omp40* in *A. ferrooxidans* ATCC 53993 grown in different copper concentrations. Error bars indicate standard deviations based on three different experimental values. Application of *t-*Student test were: \*\*\* *p* ≤ 0.001 \*\* *p* ≤ 0.01 and \* *p* ≤ 0.05.

Currently there are no efficient and easy to reproduce methods to generate knock-outs of genes in *A. ferrooxidans* [43]. Therefore, to ascertain whether the *mdoG* gene confers Cu-resistance to a heterologous host, it was expressed in *E. coli* as described in Materials and Methods. As seen in Figure 4, *A. ferrooxidans* putative *mdoG* gene conferred resistance to Cu when expressed in *E. coli* due to the increase of its MIC value from 1.0 to 3.0 mM copper. This result supports the possibility of MdoG being a copper resistance determinant in this acidophilic microorganism. In addition, the effect of overexpressing *mdoG* in *E. coli* was also tested in cells grown in the presence of Zn or Ni as shown in Figure 4.

**Figure 4.** Heterologous functional analysis of the overexpression of *A. ferrooxidans mdoG* (*Lferr\_1075*) gene in the Cu-sensitive *Escherichia coli* (K-12 Δ*copA/*Δ*cusCFBA/*Δ*cueO*) grown in zinc (Zn), nickel (Ni) and copper (Cu). pTrcHis2A, empty vector; pTrcHis2A\_1075 contains *mdoG* gene. Error bars indicate standard deviations based on three different experimental values.

Compared with copper, MdoG did not confer tolerance to Ni and Zn. Interestingly, a proteomic analysis of the response of *Rhodobacter sphaeroides* to high cobalt concentrations has been reported [44]. One of the changing proteins in presence of the metal was MdoG. It was previously suggested that cobalt would generate an alteration of permeability of the envelope, periplasm or cell wall as a possible resistance mechanism in this microorganism [44]. Whether the effect of copper is rather specific for MdoG from *A. ferrooxidans* remains to be elucidated. Another interesting protein found to be overexpressed in presence of copper was *Lferr\_0408* (Wzy) (Table 1), a protein involved in O-antigen biosynthesis, the most external segment of LPS [45,46]. By expressing gene *wzy* from *A. ferrooxidans* in the Cu-sensitive *E. coli* (K-12 Δ*copA/*Δ*cusCFBA/*Δ*cueO*) strain already used for *mdoG* gene, the results seen in Figure 5 were obtained. Once again, it is clear that expressing *wzy* gene in the heterologous host confers it a higher copper resistance.

**Figure 5.** Heterologous functional analysis of overexpression of *A. ferrooxidans* antigen-O polymerase gene *wzy* (*Lferr\_0408*) in Cu-sensitive *E. coli* (K-12 Δ*copA/*Δ*cusCFBA/*Δ*cueO*) grown in copper. pTrcHis2A, empty vector; pTrcHis2A\_0408 contains gene *wzy*. Error bars indicate standard deviations based on three different experimental values.

To support this result, the analysis of transcriptional expression of this and other genes involved in LPS generation was carried out. Figure 6 shows the levels of transcriptional expression of *wzy, wzz* and *rfaE* genes, all involved with LPS synthesis, in both *A. ferrooxidans* strains grown in the absence or presence of copper.

**Figure 6.** Transcriptional levels of genes *wzy, wzz* and *rfaE* related to lipopolysaccharides (LPS) synthesis in *A. ferrooxidans* exposed to copper. (**A**) Strain ATCC 23270; (**B**) Strain ATCC 53993. Values were obtained from three biological replicates. Error bars indicate standard deviations based on three different experimental values. Application of *t*-Student test were: \*\*\*\* *p* ≤ 0.0001 \*\* *p* ≤ 0.01 and \* *p* ≤ 0.05.

Strain ATCC 23270 clearly showed an increased level in expression of tested genes in presence of the metal (Figure 6A). On the contrary, the same genes did not show significant changes in their expression when strain ATCC 53993 was grown in presence of copper (Figure 6B). These results strongly suggest strain ATCC 23270 could synthesize higher amounts of LPS in presence of copper compared to ATCC 53993. However, when the amounts of LPS were determined in both strains in absence of the metal, ATCC 53993 showed about 3-fold higher amounts of LPS compared with ATCC 23270 (Figure 7A). This result indicates that normally, strain ATCC 53993 in addition of having extra copper resistance determinants in its genomic island, contains higher LPS levels compared with strain ATCC 23270. This could explain in part the higher copper tolerance of the former strain.

**Figure 7.** Relative LPS concentration of *A. ferrooxidans* ATCC 23270 and ATCC 53993 grown at different concentrations of CuSO4. (**A**) Cells grown in absence of copper; (**B**) Cells grown in 100 mM CuSO4. Values were obtained from three biological replicates. Error bars represent standard deviations for each condition. A *t*-Student statistic analysis was performed, where: \*\*\* indicates *p* ≤ 0.001 and ns indicates *p* > 0.05.

Nonetheless, in presence of 100 mM copper both strains showed similar LPS levels (Figure 7B), suggesting that strain 23270 increases its LPS levels in presence of the metal, in agreement with results of the transcriptional expression of its genes in the presence of copper seen in Figure 6. Previously, *A. ferrooxidans* ATCC 53993 subjected to 40 mM Cu showed an increased level of protein RfaE possibly involved in LPS synthesis [22]. Apparently, LPS could bind metals in the cell surface depending on the

composition of the polymers [47]. A summary of the main results obtained is shown in the working model of Figure 8.

Remarkably, it has been reported that *A. ferrooxidans* adapted to high copper and zinc ions concentrations showed changes in the surface chemical properties of this bacterium. Under these conditions, their surface negative charge was decreased due to changes in the structure of its surface layers [48].

**Figure 8.** Summary working model of some proteins in *A. ferrooxidans* adapted to grow in presence of copper mentioned in this study. Yellow, proteins that increase their synthesis or transcript levels in presence of copper. Pink, proteins down-regulated in cells subjected to copper. OPG stands for osmoregulated periplasmic glucans. The locations and order in which proteins are illustrated are arbitrary.

### **4. Conclusions**

The results presented here clearly indicate that several envelope components from *A. ferrooxidans* such as RND efflux pumps, LPS, porins, and periplasmic protein MdoG may be of great relevance for both, copper resistance and/or tolerance in their environment. Similar roles for these components in other members of the biomining bacterial consortia are also possible and their study may be of importance for industrial bioleaching operations.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/9/7/347/s1. Figure S1: Functional categories of *A. ferrooxidans* proteins changing their synthesis levels in cells grown in presence of 100 and 200 mM CuSO4. Table S1: Functional categories and numbers of *A. ferrooxidans* ATCC 53993 proteins changing their synthesis levels in cells grown in presence of 100 and 200 mM CuSO4. Table S2: Proteins with increased levels in *A. ferrooxidans* ATCC 53993 grown in presence of 100 and 200 mM CuSO4. Table S3: Proteins with lower levels of synthesis in *A. ferrooxidans* ATCC 53993 grown in presence of 100 and 200 mM CuSO4.

**Author Contributions:** N.O., C.A.N., S.A.Á., A.P., and C.A.J. conceived and designed the experiments. N.O., R.N., F.A., and C.M.-B. performed the experiments. N.O., R.N., A.P. and C.A.J. wrote the paper. All authors revised the manuscript.

**Funding:** This research was funded by FONDECYT grant number 1150791 to Carlos A. Jerez.

**Acknowledgments:** This work was part of the Ph.D. thesis of Nia Oetiker supported by CONICYT-PCHA/Doctorado Nacional/2013- 21130264.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Genes* Editorial Office E-mail: genes@mdpi.com www.mdpi.com/journal/genes

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18