Change of bacterial communities in sediments along Songhua River in Northeastern China after a nitrobenzene pollution event

More than 100 tons of nitrobenzene and related compounds were released into Songhua River due to the explosion of an aniline production factory in November, 2005. Sediment samples were taken from the heavily polluted drainage canal, one upstream and three downstream river sites. The change of bacterial community structures along the river was studied by denaturing gradient gel electrophoresis (DGGE) and cloning and sequencing of 16S rRNA genes with ﬁve clone libraries constructed and 101 sequences acquired representing 172 clones. Both DGGE proﬁles and sequences of 16S rRNA genes from clone libraries demonstrated that the contaminated drainage canal and three downstream river sites were similar in that all had Betaproteobacteria , mainly grouped into Comamonadaceae , as the dominant group of bacteria, and all had Firmicutes , primarily as Clostridium spp. These results suggest that these latter two groups of bacteria may play potential roles in degradation and detoxiﬁcation of nitrobenzene in the present contaminated river environments.


Introduction
Nitrobenzene is mainly used in the manufacture of aniline as the primary starting material. It is frequently released into the environment from effluent of plants of explosives, organic chemicals and plastics. Owing to its toxicity, nitrobenzene has been listed as a priority pollutant by the United States Environmental Protection Agency (Padda et al., 2003;US EPA, 2006). Numerous bacteria able to degrade nitrobenzene have been isolated. These include Acidovorax sp. strain JS42, Comamonas sp. strains JS765 and CNB-1 and Pseudomonas spp. from nitrobenzene-contaminated facility or waste treatment plants. Mechanisms of degradation of nitrobenzene and other nitroaromatic compounds with isolated strains have been investigated extensively Zhao & Ward, 2001;Parales et al., 2005;Wu et al., 2006). It has been found that the removal of nitro groups from nitroaromatic compounds by microorganisms may take place via oxidative pathways with monooxygenases or dioxygenases (Nishino & Spain, 1993, 1995Lessner et al., 2003), or reductive pathways with nitroreductases, yielding nitroso, hydroxylamino or amino derivatives (Marvin-Sikkema & de Bont, 1994;Somerville et al., 1995;Spain, 1995). Biodegradation of nitrobenzene in a lab-scale reactor has also been investigated (Dickel et al., 1993); however, to our knowledge, few studies have been performed focusing on bacterial communities in nitrobenzene-polluted environments.
Songhua River is located in the northeast part of China. With a length of about 1840 km, it flows generally northand eastward, crossing Jilin and Heilongjiang provinces, to Amur River and to the Sea of Okhotsk. The river is the main source of water for cities and villages along the river. On 13 November 2005, more than 100 tons of nitrobenzene and related compounds were spilled into Songhua River through a drainage canal due to the explosion of a chemical plant located in Jilin City, Jilin Province. The concentration of nitrobenzene in river water reached c. 0.5 mg L À1 at c. 400 km downstream from the discharging point. Concern has been expressed about comprehensive environmental impacts of the toxic spill on the region's ecosystem (World Health Organization Department for Health Action, 2005). Considering the important role of bacteria in the degradation of chemical compounds in the environment, determination of bacterial communities in sediments along the severely contaminated river should provide some clues about the response of bacterial communities to the nitrobenzene contamination and potential functionally important bacterial groups in a nitrobenzene-polluted environment. DGGE fingerprinting and cloning and sequencing of PCR-amplified 16S rRNA gene fragments have been successfully applied for the analysis of bacterial community structures in a wide range of environmental samples (Lipson & Schmidt, 2004;Webster et al., 2004;Hobel et al., 2005). In this study, these two approaches were used to determine the change of bacterial community in the sediments sampled along nearly 400 km of the contaminated river to gain an insight into the environmental influence of the toxic spill.

Study site and sampling
The main pollutant spilled into the river was nitrobenzene. Some related compounds like benzene, aniline and xylol were also discharged in small amounts. To determine the environmental influence of nitrobenzene contaminations, sampling campaigns were performed on 24 December 2005, about 40 days after the explosion. The sampling site A was located upstream of the discharging point as a reference, the contaminated site B was at the drainage canal and sites C-E were at downstream of the river. Sites C-E were all chosen at the turnings of the river where pollutants might accumulate. Between these sites, there were no cities or towns by the side of the river, as well as other pollution sources. The velocity of river flow ranged from 0.5 to 1 m s À1 during the sampling period. Sediment samples of each site were taken in triplicate and pooled, using a grab bucket after breaking the ice layer covering the river. Then sediments were transported on ice to the laboratory and maintained at À 20 1C immediately. About 4 L of water samples were also taken from each sampling location for nitrobenzene analysis using liquid chromatography MS. The HPLC system consisted of an Alliance liquid chromatograph 2695 (Waters, Milford, MA) and MS was carried out with a single-quadruple mass spectrometer ZQ 4000 (Micromass, Manchester, UK). The details of the location of sampling sites and nitrobenzene concentrations in sediments and water samples are listed in Table 1.

Batch anaerobic incubation experiment
For each sampling site, 4-g sediment samples in dry weight were added to a 22-mL airtight glass bottle, and filled with sterile mineral media spiked with 11 mg L À1 of nitrobenzene. The bottle was then sealed with a Teflon-Silica gel stopper and covered with a parafilm. Two parallel bottles and one blank sterilized with 1% HgCl 2 (w/w) were set up in a double-layer airtight plastic bag with one anaerobic package (AnaeroPack, Mitsubishi Gas Chemical Co., Japan). The bottles were incubated at 10 1C without shaking. Residual nitrobenzene concentrations were determined at a given time interval.

DNA extraction and PCR amplification
DNA was extracted from 500 mg of sediments using the method of Tsai & Olson (1991). About 15 mg DNA g À1 of wet sediment was obtained using electrophoresis on 1% (w/v) agarose gel and comparison with a molecular mass ladder visually. To acquire suitable amplicons, 10-100-fold dilutions of crude DNA were used as templates for subsequent PCRs.
For DGGE analysis, the V3 region of 16S rRNA gene was amplified using touchdown PCR methods described by Muyzer et al. (1993). The standard 50-mL PCR mixture (Takara, Dalian, China) included 1 Â PCR buffer containing 1.5 mM MgCl 2 , 200 mM of each deoxynucleoside triphosphate, 10 pmol of each primer, 1.25 U of TaKaRa rTaq polymerase and about 40 ng of template DNA. Amplification products were confirmed by electrophoresis in 1.5% (w/v) agarose gel. Meanwhile, almost full-length 16S rRNA gene sequences were amplified using universal primers 27F (5 0 -AGAGTTTGATCCTGGCTCAG) and 1392R (5 0 -GACG GGCGGTGTGTAC) (Brofft et al., 2002). The amplification reaction mixture was the same as above, and the conditions were: 95 1C for 10 min, 30 cycles of 95 1C for 1 min, 55 1C for 1 min and 72 1C for 1 min 30 s, and a final extension at 72 1C for 30 min. Amplicons were purified with the Qiaquick PCR cleanup kit (Qiagen Inc., Chatsworth, CA). To minimize PCR bias, three separate reactions were run for each sample and pooled.

DGGE analysis
DGGE was performed on a D-Code apparatus (Bio-Rad, Hercules, CA) under the same conditions as described w ND, not determined. The detection limits of the nitrobenzene concentration were 0.2 mg L À1 for water samples and 0.05 mg kg À1 for sediment samples. (Muyzer et al., 1993), with 30 mL of PCR products loaded. Gels were stained with SYBR Green I and visualized using UV transillumination (Gel Doc 2000, Bio-Rad Laboratories; Milan, Italy). Individual band patterns were compared with each other using the pairwise similarity coefficient of Dice (S) as follows: S = 2j/(a1b), where a is the number of DGGE bands in pattern 1, b is the number of bands in pattern 2 and j is the number of common bands. Then cluster analysis of band patterns was performed using the unweighted-pair group method using average linkages (UPGMA).

Cloning and sequencing of 16S rRNA genes
The cloning of amplified 16S rRNA gene fragments into the TOPO TA cloning vector pCR2.1 and the selection of TOP10 Escherichia coli transformants were all performed following the manufacturer's instructions (Invitrogen, Shanghai, China). Cloned inserts were amplified from lysed colonies with vector-specific primers M13F (5 0 -GTAAAAC GACGGCCAG) and M13R (5 0 -CAGGAAACAGCTATGAC). PCR products were digested (3 h, 37 1C) with HaeIII (Takara) and separated by electrophoresis in 2% agarose gels, and then grouped according to restriction fragment length polymorphism (RFLP) patterns. Representative clones were sequenced using an ABI 3730 automated sequencer (Invitrogen).

Phylogenetic and statistical analysis
After editing and trimming manually using BioEdit (Hall, 1999), DNA sequences were searched against the Ribosomal Database Project II (RDP II) release 9.49 (Cole et al., 2007) and the GenBank database using the BLASTN program (Altschul et al., 1997). The most similar reference sequences were retrieved and aligned with clone sequences using CLUSTALX (Thompson et al., 1997). Phylogenetic trees were constructed using MEGA version 3.1 by the neighbor-joining algorithm and the Jukes-Cantor distance estimation method with bootstrap analyses for 1000 replicates (Kumar et al., 2004). Possible chimeras were checked using CHIMERA_CHECK in RDP II and the software BELLEROPHON (Huber et al., 2004). The sequences sharing 97% or greater similarity were grouped into the same operational taxonomic unit (OTU) using software DOTUR (Schloss & Handelsman, 2005). Coverage (C) was calculated as follows: C = 1 À (n 1 /N), where n 1 is the number of OTUs that occurred only once in the clone library and N is the total number of clones (Singleton et al., 2001). OTU richness S Chao1 and S ACE were calculated using the software ESTIMATES version 8.0 with 100 random sample repetitions (Colwell, 2005). OTU diversity and distribution in the library was evaluated using Shannon diversity (H) and evenness (E) indices. Rarefaction curves were constructed using the software ARAREFACTWIN available at http://www. uga.edu/$strata/software.html. UniFrac computational analysis was performed to compare libraries from different sites following the software instructions (Lozupone & Knight, 2005). Bacterial communities from individual sites were clustered by application of the UPGMA method to the UniFrac metric matrix. Principal components analysis was performed with UniFrac metric matrix.

Nucleotide sequence accession numbers
The 101 16S rRNA gene sequences were submitted to the GenBank database under accession numbers EF589963-EF590063.

Results and discussion
Nitrobenzene residues and nitrobenzene degradation activity in the sediment The concentrations of nitrobenzene residues in sediment and water samples about 40 days after the nitrobenzene pollution accident are shown in Table 1. Nitrobenzene was only detected in sediments and water samples from sites B and C, demonstrating that nitrobenzene might be degraded in sediments. This was further confirmed by the results of a batch anaerobic incubation experiment at 10 1C (Zonglai et al., 2008). Nitrobenzene removal rates of contaminated site B and intermediate downstream site C were found to be higher (0.083-0.101 mg L À1 h À1 ) than those of downstream sites D and E (0.050-0.057 mg L À1 h À1 ). Nitrobenzene (11 mg L À1 ) was almost completely removed in the sediments from sites B-E in 8 days with nitrosobenzene and aniline as the main degradation intermediates. In contrast, the removal rate in reference site A was remarkably low and complete removal needed more than 30 days. The enhanced nitrobenzene-removing abilities for the sediments of contaminated site B and downstream sites C-E could be due to the influence of nitrobenzene pollution.

DGGE profile of 16S rRNA gene fragments
In this study, the change of bacterial community along river sediments was elucidated by the DGGE phylotype. As shown in Fig. 1, DGGE bands 1-5 were detected in all profiles, and several band positions were unique for some sampling sites, such as bands 6 and 7 for A, bands 8 and 9 for B and bands 10 and 11 for C and D. Meanwhile, the intensities of several bands like 1 and 3 in B, C and D were stronger than those in A, indicating that the bacterial groups represented by these bands possibly played an important role in nitrobenzene biodegradation.
DGGE patterns were compared with one another by pairwise similarity coefficient of Dice (S), which ranged from 0.551 to 1.000. Cluster analysis demonstrated that band patterns of C, D and E were closely related to one another, with S values of 0.829-0.838 and the pattern of B was similar, to some extent, to C, D and E (S values, 0.680-0.688) and significantly different from A (S value, 0.551). These results indicate that, in comparison with reference site A, bacterial communities in contaminated site B and downstream sites C-E retained some common characteristics that were possibly due to the influence of nitrobenzene pollution.

16S rRNA gene clone libraries
Furthermore, to determine the detailed compositions of bacterial communities in sediments, five 16S rRNA gene clone libraries were constructed and c. 30-40 clones were processed by RFLP for each library; at least one clone of each pattern was sequenced. After discarding three possible chimeras and one sequence of poor quality, 101 sequences of c. 1400 nucleotides were acquired (23 for A, 15 for B, 24 for C, 23 for D and 16 for E) and 88 OTUs were determined.
Using UniFrac metric analysis, clones of downstream sites C-E as well as contaminated site B were found to be significantly different from those clones of reference site A, consistent with the results of DGGE cluster analysis (Fig. 2). Coverage values (C) and Shannon diversity indexes (H) were calculated for five bacterial clone libraries. Both results indicated that the sequence populations from reference site A and intermediate downstream site C were more diverse than those from contaminated site B and downstream sites D and E. This was further confirmed by higher richness (S Chao1 and S ACE ) and evenness of sites A and C (Table 2) and was consistent with the DGGE band profile as shown in Fig. 1. However, as rarefaction curves did not reach saturation (data not shown), the clone number for each sample was not sufficient and may affect the index values. Considering that intermediate downstream site C was adjacent to reference site A, these statistical indexes, together with the DGGE and UniFrac metric analysis, demonstrated that the  nitrobenzene spill leads to a decrease in bacterial diversity in polluted river sediments.

Phylogeny of bacteria in sediments
All bacteria from the cloned libraries were classified into the phyla Proteobacteria, Bacteroidetes, Acidobacteria, Actino-bacteria, Chloroflexi, Planctomycetes, Gemmatimonadetes and Verrucomicrobia (Figs 3 and 4), all of which were commonly detected in freshwater sediment environments as reported previously (Tamaki et al., 2005;Wilms et al., 2006;Winter et al., 2007). Clones belonging to Proteobacteria, Bacteroidetes and Acidobacteria were present in all river sites. These phyla have been usually identified as dominant groups in sediment environments and were generally important contributors to biogeochemical processes (Barns et al., 1999;Spring et al., 2000).
The clones from reference site A were mainly grouped into the classes Acidobacteria, Alphaproteobacteria and Actinobacteria. The most notable characteristics for clones of contaminated and downstream sites B-E were the predominance of Betaproteobacteria and Firmicutes. Until now, many reported nitrobenzene-degrading bacteria are Comamonadaceae of the class Betaproteobacteria (Groenewegen & de Bont, 1992;Schenzle et al., 1999;Zhao & Ward, 1999). The clones belonging to Comamonadaceae were indeed similar to the reported nitroaromatic compounds-utilizing strains. As listed in Table 3, clone OTU B24 from contaminated site B was 99.6-99.8% similar to the nitrobenzene-degrading Acidovorax sp. strain JS42 (Lessner et al., 2003) and pchloroaniline-degrading Diaphorobacter sp. strain PCA039. Another OTU B02 showed 97-98% similarities to several strains of Comamonas sp. including phenol or benzoic aciddegrading PND-3, PP3-1 (accession no. EU276094) and   Jeon et al., 2003) and Polaromonas sp. strain CJ2 responsible for naphthalene biodegradation. A high similarity has been observed between nitrobenzene and naphthalene degradation gene clusters (Lessner et al., 2002). These results indicated that the genera Comamonas, Acidovorax, Diaphorobacter, Rhodoferax and Polaromonas of the family Comamonadaceae possibly played an essential role in nitrobenzene degradation of this study. Sequences of the phylum Firmicutes were mostly classified into Clostridium spp. and showed a relationship with Clostridium acetobutylicum and Clostridium thermoaceticum. It has been reported that C. acetobutylicum and C. thermoaceticum could reduce the nitro group of nitroaromatics to the corresponding amines (Rafii et al., 1991;Gorontzy et al., 1993;Khan et al., 1997;Huang et al., 2000). These results suggest that Clostridium may play a role in reduction of the nitro group of nitrobenzene in polluted sediments of this study.
Several sequences in the class Gammaproteobacteria from contaminated site B and downstream sites C-E also showed similarities to the strains able to utilize nitroaromatic compounds. For example, the sequence C18 from intermediate downstream site C was related to phenol-degrading Acinetobacter sp. strain PHD-4 and B07 from contaminated site B was similar to 1,2,4-trichlorobenzene-degrading Pseudomonas stutzeri strain T7 (98%). In addition to Comamonadaceae of the class Betaproteobacteria, numerous Pseudomonas spp. strains have also been reported to be capable of degrading nitrobenzene (Nishino & Spain, 1993). However, in this study, only one sequence B07 from contaminated site B was related to Pseudomonas spp.
It is interesting that many clone sequences were also similar to those obtained from cold environments. For example, the OTU A23 in the Bacteroidetes showed similarity to the bacterium FJS32 (96.8%) isolated from subglacial sediments in Southern Hemisphere glaciers (Foght et al., 2004), two OTU sequences C09 and D15 in this phylum were closely related to Antarctic bacterium R-7550 (97.3-98.5%), and two OTU sequences C03 and C12 in the class Betaproteobacteria were related to Rhodoferax spp. clones BFM 9H and KAR67 derived beneath a high Arctic glacier (Van Trappen et al., 2002;Cheng & Foght, 2007;Hansen et al., 2007). These results might be related to the fact that the river is covered by ice for nearly 5 months annually and the sediment temperature was below 4 1C at the time of sampling, suggesting that cold environments favor the growth and survival of specific organisms on a worldwide basis. The results of this study indicated that the nitrobenzene spill has significantly influenced the bacterial communities in the river sediments. Strains belonging to Comamonadaceae and Clostridium spp. may play a role in degradation and detoxification of nitrobenzene in contaminated river environments.