Phylogenetic and functional diversity of metagenomic libraries of phenol degrading sludge from petroleum refinery wastewater treatment system

Sampling

Sludge samples were collected from two different refinery wastewater treatment plants of the petroleum producer Petrobras (Brazil). The first sludge sample, named MBR1 and originated from a laboratory-scale membrane bioreactor (2 L) (MBR), was performed with continued aeration and collected after a 30-day period of high phenolic load feeding (68.5 mg L^-1), as previously described by Viero and collaborators (Viero et al. 2008).

The second sample, named MBR2, was collected from a pilot submerged aerobic membrane bioreactor, previously described by Silva et al. (2010b), that has been in continuous operation for 18 months and is part of the industrial WWTP of Petrobras Refinery (Brazil). However, a phenol acclimation step was done to increase the possibility of finding genes related with phenol degradation. The sludge sample MBR2 was acclimatized in batch-culture for a 30 day-period to 1.0 g L^-1 of phenol (Merck, USA) and then used for the metagenomic library construction. The acclimation step was performed, in triplicate, using 2.0 g L^-1 of sludge, which was centrifuged and placed into an Erlenmeyer flask containing 300 mL of an initial rich nutrient medium (2.75 g L^-1 K₂HPO₄, 2.25 g L^-1 KH₂PO₄, 0.1 g L^-1 NaCl, 1.0 g L^-1 (NH₄)₂SO₄, 0.2 g L^-1 MgCl₂.6H₂O, 0.01 g L^-1 CaCl₂ and 1 g L^-1 yeast extract as carbon source). These flasks were incubated at ambient temperature at 150 rpm. The initial carbon source was gradually diminished and replaced with phenol, in the proportion of 0.5 g L^-1 less of yeast extract for each 0.2 g L^-1 increment of phenol, until the yeast extract was totally eliminated. The modification of the initial medium composition was performed every three days, where an aliquot of the previous culture medium (10%) was added as inoculum to a new culture medium containing more phenol. The acclimatizing process was monitored through phenol quantification using gas chromatography. The sludge was collected after the microorganisms were considered totally adapted to 1.0 g L^-1 phenol, when 100% phenol was removed in less than 24 hours.

Nucleic acid extraction and metagenomic fosmid library construction

High molecular weight DNA extraction from sludge samples was carried out using the protocol previously described by Silva et al. (2010a). Fosmid libraries were constructed using the CopyControl™ HTP Fosmid Library Production Kit (Epicentre, USA), according to the manufacturer's instructions. The DNA fragment size of interest was selected prior to the construction of the metagenomic libraries. The extracted DNA was run on a 1% low-melting agarose gel (Sigma, USA) and submitted to pulsed field gel electrophoresis (Pulsed-field CHEF DRIII System - BioRad- USA) at 0.5 s switch time, 9 Vcm^-1, 120° included angle, 5 h at 14°C. DNA fragments of about 25-50 kb were isolated from the agarose gel, submitted to the end-repair reaction and ligated into the pCC2Fos fosmid vectors. These vectors were packed and transfected into Escherichia coli EPI300TM-T1^R cells, which were plated onto Luria-Bertani (LB) agar medium containing chloramphenicol (12.5 μg/mL) and incubated overnight at 37°C. Two metagenomic libraries were constructed, one for each sludge sample (MBR1 and MBR2).

For the validation of the metagenomic libraries, six fosmid clones were randomly selected from each library. The fosmid DNA of each clone was extracted using the FosmidMax DNA Purification kit (Epicentre, USA), according to the manufacturer's protocol, and then digested using 10 U NotI (Promega, USA) at 37°C overnight. The band profiles of fosmid clones were checked in preparative pulsed field gel electrophoresis (Pulsed-field CHEF DRIII System - BioRad- USA) at angle 120°, 6 Vcm^-1, 1 s - 12 s switch time, 10.5 h at 14°C.

Extraction of fosmid DNA pools and pyrosequencing

Clones derived from both metagenomic libraries were grown in 96-well plates containing 1 mL of LB medium and chloramphenicol (12.5 μg/mL) for 17 h at 37°C at 180 rpm. For fosmid extractions, an aliquot of 500 μL was aliquoted from each well of ten 96 well-plates and pooled, totaling a final volume of approximately 500 mL for each extraction. Each pool, containing 960 clones, was extracted using the Large-Construction Kit (Qiagen, USA), according to the manufacturer's protocol.

Fosmid DNA extracted from all pools for both libraries was then combined to produce a single DNA sample, a master pool of 5 μg DNA, for subsequent pyrosequencing. This master pool was firstly nebulized, for 2 minutes, to produce DNA fragments with average size of 500 bp. The DNA fragment ends were polished and purified to remove fragments < 500 bp and the Roche 454 GS FLX Titanium sequencing (454 Life Sciences Branford, CT, USA) was performed according to 454/Roche GS-FLX protocols.

Assembly and sequence analysis

The sorting and trimming of the metagenomic data, based on quality and size of the reads, as well as the contig assembling were done using the 454 Newbler assembler (version 2.0.01.14) for Genome Sequencer FLX (Roche, USA).

For the phylogenetic classification, the metagenomic data were compared using BLASTn (E-value ≤ 1e-2) against all 16S rRNA sequences deposited in the European Ribosomal RNA Database (http://bioinformatics.psb.ugent.be/webtools/rRNA/). Then, all reads present in the metagenomic data that matched with any rRNA sequence were selected and submitted to a phylogenetic classification using the RDP naïve Bayesian rRNA Classifier tool, at 80% confidence threshold, and E-value of ≤ 1e-2, available on the Ribosomal Database Project's Pyrosequencing Pipeline website.

The contig and singlet sequences were analyzed with MEGAN software and the MG-RAST platform. These sequence sets were submitted to protein BLAST for sequence matching using the NR database (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The resulting BLASTp file was then analyzed using MEGAN metagenomic software (Huson et al. 2007). MEGAN software uses a homology-matching algorithm to generate a phylogenetic tree based on the NCBI taxonomic database.

The metabolic pathway classification of contig and singlet sequences was done using the MG-RAST platform (http://metagenomics.nmpdr.org/metagenomics.cgi?page=Logout). The data were submitted as text file (.txt) for annotation using the subsystems technology. In this approach, reads are classified in a hierarchical structure in which all genes required for a specific function are grouped into subsystems (Aziz et al. 2008). An overview of the experimental approach used in this study is illustrated in Figure 1.

Nucleotide sequence accession number

The raw data set from the fosmid libraries is available from the NCBI Short Read Archive (SRX038779.2).

GC-MS analysis of phenol of acclimated sample

To verify if the MBR2 sample microorganisms were degrading phenol after acclimation process, the phenol concentration was determined by 1:1 extraction with CH₂Cl₂ according to the procedures described in Standard Methods (1998). Phenol quantitation was done using a calibration curve constructed from a fortified matrix in three levels. Growth culture samples were filtered through a 0.45 μm membrane filter to remove suspended biomass before analysis. One microliter of extracted sample was subsequently injected into a gas chromatography/mass spectrometry (model 6890 N, Agilent, USA) equipped with a 5975 selective mass detector and 7683B automatic injector. The sample components were separated in the HP-5 MS (30 m × 0.25 mm × 0.25 μm) capillary column using the following temperature programming: 60°C (3°/min), 240°C/10 min. The injector temperature was 220°C and the interface temperature was 250°C. Split ratio was 1/100 and the monitoring fragments were 94 (quantification ion), 66 and 65 (identification ions).

MBR2 sludge acclimation

The acclimation step was done only for the MBR2 sample, since this sludge sample originated from a membrane bioreactor operating without high phenol concentration. The biomass was progressively acclimated to a maximum phenol concentration of 1.0 g L^-1 for 30 days, and after this period the biomass was able to degrade 1.0 g L^-1 of phenol in less than 24 hours.

The biomass acclimation was checked by monitoring phenol degradation using visual and chromatographic analysis during 30 days. The visual analysis consisted of observation of microbial aggregates (flocculation) and the production of yellow pigments in the culture medium, which, according to Harayama et al. (1992), is indicative of phenol biodegradation through the meta-cleavage pathway. The yellow color comes from the product of cathecol cleavage reaction, 2-hydroxymuconic semialdehyde. Additionally, the phenol degradation was monitored by chromatographic experiments, which demonstrated that 90% of phenol was degraded in the first 12 hours, and in less than 24 hours 1.0 g L^-1 of phenol was totally removed by the acclimated biomass.

Metagenomic fosmid libraries and pyrosequencing

Two metagenomic fosmid libraries were made using high molecular weight DNA extracted from the total microbial community of MBR1 and MBR2 sludge samples, yielding 10,000 and 3,200 clones, respectively. The libraries were validated by enzyme digestion in order to check the diversity of clones obtained. Pulsed field gel electrophoresis revealed distinct band profiles for the six randomly selected clones from each library, with the vector band present in the same position for all clones, confirming the occurrence of different insertion events.

The 13,200 clones obtained from both libraries were then submitted to fosmid DNA extraction in pools and subsequent pyrosequencing. A single pyrosequencing run was done for two libraries, yielding 322,742 reads with an average sequence length of 263 bp, the largest sequence being 735 bp. The metagenomic data was assembled using Newbler (version 2.0.01.14), which employs algorithm attempts to combine individual sequence reads into longer contigs. The assembly recruited 108,000 sequences, which resulted in 22,267 contigs and 99,786 singletons.

Phylogenetic composition of phenol degrading sludge by MEGAN

Community taxonomic analysis of metagenomic data was performed using MEGAN (Metagenome analysis software). This software generated a phylogenetic tree based on the NCBI taxonomic database, in which the size of each circular node is proportional to the number of assignments at the particular taxonomic level. The phylogenetic tree showed that the majority of the metagenomic sequences were affiliated to the Bacteria domain (64%) and the remaining sequences were distributed between the Archaea and Eukarya domains (Figure 2). The eukaryotes were represented mainly by fungi and protozoa, common organisms present in the sludge of wastewater treatment system (Jiang et al. 2005; Oliveira et al. 2009). A considerable percentage of metagenomic sequences (34%) were unaffiliated. Possible these sequences are distantly related with any known sequence deposited in the public databases.

The Bacteria domain was represented by sixteen different phyla encompassing Proteobacteria, Candidate Division WWE1, Bacteroidetes/Chlorobi, Planctomycetes, Firmicutes, Deinococcus-Thermus, Fusobacteria, Chloroflexi, Cyanobacteria, Actinobacteria, Thermotogae, Synergistetes, Acidobacteria, Verrucomicrobia, Lentisphaerae and Gemmatimonadetes (Figure 2). Among these, the most abundant were Proteobacteria (66%), followed by Bacteroidetes (7%), Actinobacteria (5.5%), Planctomycetes (4.9%) and Verrucomicrobia (1.3%). This finding supports previous reports of our research group that proteobacteria is a predominant phylum in sludge samples (Silva et al. 2010a; Silva et al. 2010b), suggesting that these organisms play key roles in biodegradation processes that take place in petroleum refinery wastewater treatment systems.

The metagenomic data analysis using MEGAN also yielded a profile of attributes for the bacterial community, in which the majority of the bacterial species were shown to be gram-negative, aerobic, motile, non-pathogenic, aquatic and mesophilic (Figure 3). These characteristics are in accordance with those displayed by the vast majority of proteobacteria detected in the sludge of WWTPs (Miura et al. 2007; Kraigher et al. 2008; Sanapareddy et al. 2009, Silva et al. 2010a, Silva et al. 2010b).

Phylogenetic classification of metagenomic sequences

The comparison of all metagenomic sequences with the European Database allowed the selection of 6,099 sequences presenting some 16S rRNA gene similarity, this dataset was submitted to automatic annotation using the RDP Classifier. This algorithm platform uses Bayesian statistics to assign 16S rRNA genes sequences to known taxa. A set of 247 reads was annotated at least to the phylum level based on known 16S rRNA genes sequences from RDP database at an E-value cutoff ≤ 0.01. The selected size of sort reads in the pyrosequencing data was 100 nt at least. This sequence length was chosen based on the work of Liu et al. (2007) showing that reads as short as 100 bases are long enough to accurately characterize taxa. The annotation results revealed the presence of six different phyla in the sludge samples under study, named Proteobacteria, Actinobacteria, Chloroflexi, Planctomycetes, Verrucomicrobia and Bacteroidetes (Figure 4). Despite the fact that this analysis uses only 16S rRNA gene sequences for the classification, most of the results were consistent with the MEGAN analysis (Figure 2).

At the phylum level, the observed taxa were dominated by representatives of Proteobacteria, which accounted for 91% of the classified sequences (Figure 4a). From these, about 45% were affiliated to class Betaproteobacteria (Figure 4b). The majority of the 16S rRNA sequences were classified at the order level and, distributed across 11 different orders with Burkholderiales (Betaproteobacteria) the most abundant, followed by Rhizobiales (Alphaproteobacteria) and Pseudomonadales (Gammaproteobacteria) (Figure 4c). At the genus level, the sequences could be classified in 17 known genera of which Diaphorobacter, Pseudomonas, Hyphomicrobium, Comamonas, Planctomyces and Thauera were the most prevalent. As classification of the sequences moved from phylum to genus, fewer sequences could be classified with a RDP confidence score of at least 80%, and 67% of the sequences were considered unassigned (Figure 4d), confirming previous reports on the huge bacterial diversity of wastewater treatment sludges (Miura et al. 2007; Sanapareddy et al. 2009; Silva et al. 2010a; Silva et al. 2010b).

Functional analysis of the metagenome phenol degrading sludge

The sequences were functionally annotated based upon MG-RAST plataform using the SEED database. The functional analysis allowed the classification of 42.8% of the metagenomic sequences into several subsystems with the majority of them related to the metabolism of carbohydrates, aromatic compounds, proteins, DNA, amino acids and derivatives, virulence, respiration and cofactors. By contrast, some subsystems were under-represented in the metagenomic data, including sequences involved in photosynthesis, secondary metabolism, macromolecular synthesis and nitrogen metabolism (Figure 5). Despite of the low number of reads assigned to the nitrogen metabolism, the majority of these sequences represented genes coding enzymes linked to processes such as nitrate and nitrite ammonification and denitrification, which are considered important processes for the removal of complex nitrogen compounds in wastewater treatment. An interesting result was observed for the virulence subsystem (Figure 5), which accounted for a high number of reads. Annotation of such reads revealed that almost 50% of them were related to antibiotic and toxic compounds resistance genes, which are extremely important features for microbial survival and adaptation in polluted environments.

Analysis of metagenome for phenol degradation

The phenol can be aerobically degraded for two different pathways, ortho- or meta-pathway. The aromatic ring is initially monohydroxilated in the adjacent carbon of a hydroxyl group by the enzyme phenol hydroxylase (EC 1.14.13.7) resulting in catechol, which is in turn cleaved by either ortho- or meta-cleavage pathway. In case of the ortho-pathway, the ring is cleaved by the catechol 1,2-dioxygenase enzyme (EC 1.13.11.1), leading to the initial formation of succinyl-CoA and acetyl-CoA. In the meta- pathway, the catechol is cleaved by the catechol 2,3-dioxygenase enzyme (C23O), leading to the formation of pyruvate and acetaldehyde (Marimaa et al. 2006).

In MG-RAST analysis, the aromatic compound utilization profile of phenol degrading sludge was dominated by proteins annotated as subsystems of peripheral pathway for catabolism of aromatic compounds and metabolism of central aromatic intermediates, besides benzoate degradation, cresol degradation and toluene 4-mooxygenase, that are also important pathways involved in the catabolism of pollutant compounds found in refinery wastewater (Figure 6).

The subsystem of peripheral pathway for catabolism of aromatic compounds encompassed the proteins involved with the degradation of phenol and derivatives, such as phenol hydroxylase (EC 1.14.13.7) (78 reads) and enzymes involved in the biphenyl degradation (75 reads). The subsystem of metabolism of central aromatic intermediates, although mainly represented by the homogentisate pathway, involved in the catabolism of aromatic rings, contained the proteins involved in the ortho- and meta-pathway of phenol degradation, such as catechol 1,2-dioxygenase (EC 1.13.11.1) (9 reads) and catechol 2,3- dioxygenase (EC 1.13.11.2) (23 reads), respectively.

These analyses showed that the genes related to phenol degradation have been well sampled in the phenol degrading sludge, revealing that the phenol enrichment was successful. Further studies will be carried out to verify if these genes represent new sequences coming from uncultivated microorganisms.

Analysis of the metabolic potential for organic matter removal

This analysis was done to investigate the presence of some important functions related to the biological treatment performance in the microbial sludge under study after the phenol enrichment.

The potential metabolic functions for organic matter removal were established by comparison of all pyrosequencing-derived sequences from phenol degrading sludge to the KEGG maps using functional assignments from SEED analysis (Meyer et al. 2008). Within the KEGG categories, matches were separated into different subcategories and several sequences were assigned to more than one subcategory. The most represented categories were metabolism of energy, amino acids, carbohydrates, cofactors and nucleotides. These categories showed a high number of distinct Enzyme Commissions (EC) belonging to a specific metabolic pathway. The EC number is the common name for IUBMB's Enzyme Nomenclature System, and plays a key role in classifying enzymatic reactions and linking the enzyme or protein genes to reactions in metabolic pathways (Yamanishi et al. 2009).

In addition, the analysis of the carbohydrate metabolism category showed a high number of matches (9,387 sequences) related to genes involved in several subcategories for utilization of carbohydrates (Table 1). These results demonstrated the broad capability of microorganisms from refinery sludge to metabolize distinct sugar sources.