Five sampling sites had been chosen in the Iron Quadrangle region (Minas Gerais state, Brazil), which is extremely rich in ores and has been the scene of gold extraction since colonial times, and of mining for iron and manganese since the nineteenth century. Three sampling sites were chosen in mining-impacted streams, i.e., in the Mina stream (MS; 19°58′46.80″S and 43°49′17.07″W), in the Tulipa stream (TS; 19°59′08.1″S and 43°28′15.2″W), and in the Carrapatos stream (CS; 19°58′15.4″S and 43°27′50.7″W), and two sites in a non-impacted stream (19°59′12.1″S and 43°29′27.5″W) which did not experience mining activities. The sampled sites in the non-mining-impacted stream were designed S1 and S2. With the exception of the Mina stream, the streams do not have formal names. Two of them are popularly known as Tulipa and Carrapatos (Figure S1). An interesting characteristic of these streams is their circumneutral pH. Although the dissolution of minerals can cause the production of acidic wastes and lead to lower pH values (Smedley and Kinniburgh, 2002), the abundance of dolomite in the ores from the Iron Quadrangle region neutralizes the acids with consequent buffering of the pH (Thomas et al., 2000; Cidu et al., 2009; Varejão et al., 2011; Reis et al., 2013).

Samples were aseptically collected in August 2012 (dry season), and taken from bulk water (5L) and billowy sediment at a depth of 15 and 5 cm below the respective surfaces, respectively. Sediment was sampled at five random spots located in line, 1 m from each other. The sediment samples were then pooled into a single sample corresponding to the study site, and transported to the laboratory within 4 h on ice and stored at −20°C until further processing. The samples were labeled by the name of the stream followed by W (water) or S (sediment), e.g., CW (Carrapatos stream water) and CS (Carrapatos stream sediment).

To assess the bulk water conditions of the streams, physicochemical variables were measured. Temperature, pH, and dissolved oxygen (DO) were measured in situ with a Hydrolab model DS5X multi-parameter probe. Total nitrogen (TN), total phosphorus (TP), nitrite (NO⁻₂-N), nitrate (NO⁻₃-N), and soluble reactive phosphorus (PO³⁻₄-P) were measured as previously described by Mackereth et al. (1978). The concentration of ammonium (NH⁺₄-N) was measured according to Koroleff (1976). The readings were performed with a Shimadzu model UV 1700 spectrophotometer. Metal and metalloid concentrations of bulk water and sediment of all the samples were determined by using inductively coupled plasma-optical emission spectrometry (ICP-OES, Optima 7300 DV, PerkinElmer).

Total DNA was extracted from all the sediment samples (10 g wet weight) and from the Carrapatos stream bulk water sample (5L) using the UltraClean Mega Prep soil DNA kit and UltraClean Water DNA kit (MoBio Laboratories), respectively, according to the manufacturer's instructions. Quantification and quality of total DNA were determined using the equipment Agilent 2100 Bioanalyzer, again according to the manufacturer's instructions.

Wang et al. (2009) successfully applied 16S rRNA gene-based approaches using Gallionellaceae-related primers, to describe the FeOB community structure. PCR-DGGE of 16S rRNA gene was performed by a nested PCR using first the Gallionellaceae-related primers set 122F (5′-ATATCGGAACATGTCCGG-3′) and 998R (5′-CTCTGGAAACTTCCTGAC-3′) (Wang et al., 2009), followed by nested PCR using the primers set 341F/GC (5′-CCTACGGGAGGCAGCAG-3′) and 907R (5′-CCGTCAATTCMTTTGAGTTT-3′), specific for the Bacteria domain (Muyzer et al., 1993), according to Wang et al. (2011). The final PCR products were separated by DGGE in a 6% polyacrylamide gel with a vertical gradient of 30–60% of formamide and urea denaturants. The running conditions were 80 V at a constant temperature of 60°C for 18 h.

The phylogenetic assignment of Gallionellaceae-related FeOB from samples was determined by excising DGGE bands, which were eluted in 20 μl of sterile MilliQ water overnight at 6°C, and sequencing, using the same primers set 341F and 907R, in the ABI Prism 3130 DNA sequencer (Applied Biosystems, Foster City, Calif.). Partial sequences were checked for quality, aligned, and edited to produce a consensus using the Linux programs Phred/Phrap/Consed (http://www.phrap.org/phredphrapconsed.html). The Bellerophon program (Huber et al., 2004) was used to detect and omit chimeric DNAs. Sequences were compared with available databases using the BLASTn search tool from GenBank (http://www.ncbi.nlm.nih.gov/) and the Greengenes Project (http://greengenes.lbl.gov/) to identify the closest relatives of the sequences.

Sequence alignment and phylogenetic relationships were inferred with ARB (Ludwig et al., 2004; Pruesse et al., 2007) using the neighbor-joining algorithm (http://www.arb-home.de). The bootstrap consensus tree inferred from 500 replicates (Felsenstein, 1985) was taken to represent the evolutionary history of the taxa analyzed.

Bands at identical positions in the DGGE gel are not necessarily derived from the same species nor do distinct bands indicate the same bacterium (Emerson and Moyer, 1997). Thus, whenever possible, we sequenced at least two bands at similar positions on the DGGE gel. The nucleotide sequences (288 bp) generated were deposited in the GenBank database with accession numbers KJ363887 to KJ363908.

For each sediment and Carrapatos bulk water sample we estimated the abundance of the 16S rRNA gene associated with Gallionellaceae-related FeOB and with the Bacteria domain using an ABI PRISM 7900HT sequence detection system (Applied Biosystems, Foster City, CA). The primer sets 628F (5′-GBMAGGCTAGAGTGTAGC-3′) and 998R (Wang et al., 2009), and 338F (5′-TACGGGAGGCAGCAG-3′) (Raskin et al., 1994) and 518R (5′-ATTACCGCGGCTGCTGG-3′) (Muyzer et al., 1993) were used for the amplification of 16S rRNA genes from the Gallionellaceae-related FeOB and the Bacteria domain, respectively. Reaction conditions for Gallionellaceae-related FeOB and the Bacteria domain were as described by Wang et al. (2011) and Reis et al. (2013), respectively. Each sample was run in triplicate and a negative control was included for each analysis. Relative frequency values of Gallionellaceae-related 16S rRNA genes were calculated for each sample (McCune and Grace, 2002). Positive controls consisted of a mix of all DNA from samples, which were amplified using the primer sets specific for Gallionellaceae-related FeOB and the Bacteria domain, respectively. The standard curves for the primer sets generated slopes of −3.39 and −3.22, respectively, and the R² values were greater than 0.99 for both curves (Figure S2). Bacteria domain qPCR exhibited a wide range of Ct-values ranging from 13 to 25. On the other hand, lower variation of Ct-values was obtained for Gallionellaceae-related FeOB (from 24 to 27) (Figure S3).

For DGGE data, the presence or absence of co-migrating bands was converted to a binary matrix (0/1). A dendrogram was calculated using Jaccard coefficient of similarity and the un-weighted pair-group method with arithmetic averages (UPGMA). Analysis of data was performed using the software PAST (Paleontological Statistics Software Package) (Hammer et al., 2001).

To correlate the distribution of samples according to environmental parameters, we performed a Principal Component Analysis (PCA). After standardization of environmental data (by subtracting the mean from each observation and dividing by the corresponding standard deviation), PCA was obtained using the rda function in the Vegan library program implemented in R (<http://www.r-project.org/>). This statistical technique is frequently used to analyze a data table representing the environmental variables of each sample. Thus, the important information is extracted from the data table and is expressed as a set of new orthogonal variables called principal components. PCA also demonstrates the pattern of similarity of the samples and the variables by plotting them as points in maps (Jolliffe, 2002).

In an attempt to characterize the environmental conditions and correlate them with the dynamics of the neutrophilic FeOB community, we analyzed the physicochemical characteristics of the bulk stream water, as well as the metal and metalloid concentrations in bulk water and in sediment samples from the five stream locations.

Although the pH values of the bulk water samples varied largely, all were less than 1 unit different from neutral (Table 1). Values of conductivity showed also considerable variation with the highest value in the MW sample, and the lowest in the SW1 and SW2 samples. Dissolved oxygen (DO) concentrations ranged from 3.62 (TW) to 12.52 (CW) mg O₂ l⁻¹. The water samples varied widely with regard to nitrogen and phosphorus concentrations. Mining-impacted streams showed much higher NH⁺₄-N concentrations than the non-impacted stream (SW1 and SW2). Moreover, NO⁻₃-N values displayed by the Mina stream were remarkably higher than the values shown by all the other streams. Mina stream showed also higher values of NO⁻₂-N. Moreover, Tulipa stream exhibited the same concentrations of NO⁻₂-N and NO⁻₃-N. Total phosphorus concentrations allowed categorization of the streams as eutrophic (Mina stream), mesotrophic (Carrapatos stream), oligotrophic (non-mining-impacted stream), and ultraoligotrophic (Tulipa stream) (Table 1).

Metal and metalloid concentrations of water and sediment samples varied widely (Table 2). The CS sample showed strikingly high concentrations of Fe, but also considerable amounts of Mn and Mg. Interestingly, Fe concentration found in the sediments from the other mining-impacted streams (TS and MS) was lower than the sediments obtained from non-impacted stream (S1 and S2). The MS was rich in As, Cu, Zn, and Mn (Table 2).

PCA analysis revealed that the mining-impacted MS, CS, and TS sediment samples were clearly separated from each other, whereas the SS1, SW1, SS2, SW2, and CW samples were closer to each other (Figure 1). The MS sample was largely determined by the first axis (PC1) of the PCA biplot with positive values for inorganic nitrogen, Zn, Cu, and As. The MW sample was also determined by PC1 axis, with positive values for conductivity, NH⁺₄-N and NO⁻₂-N. The CS, TS, and TW samples were mainly determined by the second axis (PC2) with the variables PO₄-P, Mn, Fe, DO, and temperature. The first four variables of this axis were positively correlated to the CS sample, and negatively correlated to the TS and TW samples. The opposite was observed for the TS and TW samples. The data show a clear separation between the mining-impacted and non-impacted sediment samples, with the bulk water sample CW more closely resembling the non-impacted sediments, whereas the MW and TW bulk water samples were closer to their respective sediment samples MS and TS. The first two axes of the PCA accounted for 72% of the total variance of sample distribution (Figure 1).

Because the highest concentration of Fe among the bulk water samples was derived from the Carrapatos stream, we decided to focus the molecular analyses of water samples to water sample of this stream. Most DGGE bands representing Gallionellaceae-related FeOB were unique for a given sample, only a few were shared between two samples, and none was common to all samples (Figure S4). The CS sample produced the largest number of bands, whereas the other samples exhibited the same number of bands (Figure S4). The UPGMA cluster analysis revealed a distinct bacterial community composition of the samples analyzed (Figure 2). The Gallionellaceae-related FeOB community from the MS sample was most dissimilar, whereas the communities retrieved from SS1, CW, and SS2 samples were the least dissimilar (Figure 2).

To determine the Gallionellaceae-related FeOB community composition associated with bulk water and sediments from mining-impacted and non-impacted streams, a total of 22 DGGE bands were excised from the gel, sequenced and analyzed phylogenetically. No chimeric sequence was detected.

As expected, all sequences chosen for analysis were affiliated with the Betaproteobacteria class (Figure 3). The phylogenetic tree separated the samples in a way that was similar to the dendrogram analysis presented in Figure 2. Most sequences were affiliated with the neutrophilic FeOB genera Gallionella (7 bands) and Sideroxydans (14 bands). The Sideroxydans-related sequences were scattered throughout the samples, whereas Gallionella-related sequences were exclusive to mining-impacted sediments. Moreover, the MS-4 sequence was affiliated with the Thiobacillus genus (Figure 3), which contains also some iron-oxidizing species.

To estimate the densities of Gallionellaceae-related taxa and of total Bacteria in each sample, 16S rRNA gene copy numbers were quantified by qPCR. The abundance of copies of 16S rRNA genes/g sediment or L of water, respectively, ranged from 1.16 × 10⁴ to 1.55 × 10⁹ for Bacteria, and 1.92 × 10³ to 4.42 × 10⁵ for Gallionellaceae-related taxa (Figure S3). The TS sample obtained from an environment with low concentrations of Fe, Mn, and Ni showed the highest relative abundance of Gallionellaceae-related 16S rRNA genes with a relative frequency of 1.65 × 10⁻¹ (Figure 4). Oppositely, the CS sample obtained from the site with high concentrations of PO³⁻₄-P, Mn, and Fe showed the lowest relative abundance of Gallionellaceae-related16S rRNA genes with a relative frequency of 1.92 × 10⁻⁴. The abundance of Bacteria in the CS sediment sample was three orders of magnitude higher than in the corresponding CW water sample, whereas the relative frequency of Gallionellaceae-related 16S rRNA genes was greater in the CW sample.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.