Temporal distribution of bacterial community structure in the Changjiang Estuary hypoxia area and the adjacent East China Sea

Three cruises were conducted in June, August and October, 2006, respectively, in Changjiang Estuary hypoxia area and the adjacent East China Sea. Five stations (H32, H12, H15, H22 and H26) were chosen, with the former two stations in the hypoxia zone (figure 1). Hypoxia is defined as dissolved oxygen < 2 mg l⁻¹ (Zhu et al 2011). Samples in June, August and October were named as H-1-, H-2- and H-3-, respectively. Three layers (surface, middle and bottom) were studied for every station. For every sample, 9 l seawater (for three replicates) for microbial diversity analysis were taken with Niskin bottles fixed on an SBE25 CTD (Sea-Bird Electronics).

Seawater was filtered onto 0.2 µm pore size, 47 mm polycarbonate filters (PFs) (Millipore) after pre-filtration through 3 µm pore size, 47 mm PFs. Filters (0.2 µm) for DNA extraction were then stored at −20 °C.

Zoom In Zoom Out Reset image size

For each sample, in situ measurements of the physicochemical parameters of seawater (depth, dissolved oxygen, pH, salinity and temperature) were recorded with the SBE25 CTD. Other parameters were measured in the laboratory. The Chl a concentration was determined fluorometrically using a Turner Designs fluorometer (provided by Chenggang Liu, Second Institute of Oceanography, State Oceanic Administration). Data on nitrate, nitrite, ammonium, silicate, phosphate, particulate organic carbon (POC) and dissolved organic carbon (DOC) were obtained from the State Key Laboratory of Estuarine and Coastal Research, East China Normal University.

Genomic DNA was extracted and purified using lysozyme, proteinase K and sodium dodecyl sulfate with chloroform extraction and isopropanol precipitation, following Abell and Bowman (2005). DNA extractions from three replicates were pooled for each sample.

The v3 region of the bacterial 16S rRNA gene was amplified from genomic DNA using bacterial-specific primers 341F-GC ($5 ^{\prime}\text{-}\underline{\mathrm{CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCA}} \underline{\mathrm{CGGGGGG}}\mathrm{CCTACGGGAGGCAGCAG}$-3') and 534R (5'-ATTACCGCGGCTGCTGG-3') (Muyzer et al 1993).

The PCR was performed using the following program: 94 °C for 10 min; followed by 20 touchdown cycles of denaturation (at 94 °C for 1 min), annealing (at 65–55 °C for 1 min, decreasing 0.5 °C each cycle) and extension (at 72 °C for 1 min); 15 standard cycles of denaturation (at 94 °C for 1 min), annealing (at 55 °C for 1 min) and extension (at 72 °C for 1 min), and a final extension at 72 °C for 10 min. To reduce possible inter-sample PCR variation, all PCR reactions were run in triplicate and pooled together before loading on the DGGE gel. In PCR amplification, genomic DNA of E. coli was used as a positive control and PCR mixture without a DNA template was used as a negative control.

DGGE was performed using a Dcode™ Universal Mutation Detection System (Bio-Rad, USA). PCR products were separated on a vertical gel containing 8% polyacrylamide (acrylamide:bisacrylamide ratio of 37.5:1) and a linear gradient of the denaturants (urea and formamide), which increased from 40% at the top to 65% at the bottom of the gel. Electrophoresis was performed at 58 °C in a 1×TAE buffer at 65 V for 10 h. The gel was stained with SYBR Green I and DNA bands were visualized and photographed with a Canon A640 digital camera (Canon Co., Ltd, Japan).

For sample comparison, DGGE band position and intensity in each gel track were determined using the Quantity One Program (Bio-Rad, USA). Lane background was subtracted by using a rolling disk size of 8 (Salles et al 2004, Zeng et al 2009). Band matching was performed with the tolerance and optimization setting at 1.00% (Salles et al 2004). Cluster analysis of the DGGE banding pattern was performed with the NTSYS pc-package (Rohlf 1988) using the unweighted pair group method with arithmetic mean (UPGMA), based on the presence or absence of each band (Abell and Bowman 2005).

The relationship between DGGE profiles and the seawater physicochemical properties was studied. To test whether weighted-averaging techniques or linear methods were appropriate, detrended correspondence analysis (DCA) was performed using CANOCO for Windows 4.5 (Biometris, Netherlands).

The longest gradients resulting from DCA were 1.652, 2.794 and 1.878 for the analysis based on DGGE profiles in June, August and October, respectively. These values did not indicate a clear linear or unimodal relationship (Lepš and Šmilauer 2003). We performed canonical correspondence analysis (CCA) to compare species–environment correlations.

An automated forward selection was used to analyze inter-sample distances for DGGE profiles. First, the variance inflation factor of environmental variables was calculated. Variables displaying a value greater than 20 of this factor were excluded from CCA analyses, assuming collinearity of the respective variable with other variables included in the examined dataset.

The analysis was performed without transformation of data and focus scaling on inter-sample distances. Manual selection of environmental variables, applying a partial Monte Carlo permutation test (499 permutations) with unrestricted permutation, was performed to investigate the statistical significance (Salles et al 2004, Sapp et al 2007). The marginal effects of environmental variables were selected according to their significance level (P < 0.05) prior to permutation. Ordination biplots including the environmental variables and DGGE samples (lanes) were used to explain our data.

Strong and well-isolated DGGE bands were excised from the gel, washed with 500 µl sterile water three times and then DNA was eluted into 50  µl sterile water by incubation at 4 °C overnight. After centrifugation at 11 000g for 60 s, the supernatant was used as a template for the PCR-DGGE analysis to check for band position and purity. After that, DGGE bands were re-amplified with no GC clamp forward primer. PCR products were purified and cloned into a TA vector using pMD18-T (TaKaRa, Japan). The insertion of DNA fragments with appropriate sizes was confirmed by PCR amplification with M13-47 and RV-M primers, which corresponded to both sides of the cloning site on the vector. The 16S rRNA genes from DGGE were sequenced using the primer M13-47 or RV-M in the GenScript USA Inc. (Nanjing, China).

Sequences from DGGE bands were submitted to the GenBank database with the accession numbers EF433 334–EF433 362, EF433 369–EF433 371 and EF467 882–EF467 884.

Due to the influences of Changjiang River freshwater, the surface water at the H12 station had the lowest salinity but the highest concentration of Chl a, DOC and POC (table 1). Hypoxia was observed at stations H12 and 32 during sampling in August (table 1). In June, as a result of the intrusion of the cold water mass from the Yellow Sea (Zhao 1987), the temperature of sample H15-1-B was the lowest.

DGGE profiles of bacterial community structures in the Changjiang Estuary hypoxia and its adjacent area of the ECS are shown in figure 2. The average number of recognized DGGE bands from samples in June, August and October, were 22, 11 and 21, respectively. Cluster analysis of the DGGE profiles based on the presence or absence of bands is shown in figure 3. In June, samples at station H26 formed one cluster separated from samples at stations H12/H22/H15. Samples at station H12 and H22 grouped together, and samples from the bottom layer separated from ones from the surface and middle layers at the two stations. In August, samples H26-2-S/H12-2-B/H15-2-M formed one cluster, respectively, separating from other samples with >68% similarity. In October, most samples with >74% similarity grouped into one cluster, separating from samples H32-3-S/H32-3-M and sample H12-3-S.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Forty-seven band sequences were affiliated to Gammaproteobacteria (twenty-eight band sequences), Cytophaga–Flavobacteria–Bacteroides (CFB) (twelve band sequences, mainly Flavobacteria subgroup), Deltaproteobacteria (two band sequences), Cyanobacteria (three band sequences) and Firmicutes (two band sequences) (table 2). Most of the sequences (89% of band sequences) were related to uncultured marine bacteria with high similarity (>96% similarity). The majority of sequences were from marine ecosystems, such as the inland sea of Japan, the NW Mediterranean Sea, the eastern subtropical North Pacific, the North Sea, Plum Island Sound Estuary, the Atlantic Ocean, the North Aegean Sea, the Guanabara Bay, Arctic sea ice, the Yellow Sea and Chesapeake Bay, except for three sequences from marine Alteromonas sp. KOPRI 11568 and one sequence from symbiotic associations in sea urchin, respectively.

Comparison of bacterial community structure among the samples revealed by DGGE is displayed in figure 4 (as percentages of the different groups based on the density of their correlated DGGE bands). Two bacterial clusters, Gammaproteobacteria and CFB, were the dominant groups in all samples in June, August and October. However, the two dominant clusters presented different proportions among the three months. The average percentage of Gammaproteobacteria decreased from 63% in June to 49% in August and to 22% in October whereas the average percentage of CFB increased from 29% in June to 48% in August and to 63% in October. A higher percentage of Deltaproteobacteria was observed in samples in October (8% on average) than other samples in June and August (both 1% on average). The Firmicutes cluster was observed in samples in June and October (both 6% on average) but not in August. Small numbers of Cyanobacteria band sequences were found in the three months, on average 1%, 2% and 1%, respectively. Notably, the percentages of Cyanobacteria sequences were relatively higher at stations H32 (5% on average) and H12 (3% on average) in August, up to 12% in H32-2-S (figure 4).

Zoom In Zoom Out Reset image size

The eigenvalues of the ordination analyses are shown in table 3. The sum of all eigenvalues indicated an overall variance in the dataset of 0.743, 2.509 and 0.85 in June, August and October, respectively. The sum of all canonical eigenvalues indicated the total variation explained by environmental variation, accounting for 0.621, 1.944 and 0.525 in June, August and October, respectively. Species–environment correlations in samples in three months were high (especially for axes 1, >90%), indicating a strong relationship between species and environmental variables (table 3). Concerning the variance of species data, the first axis explained 33.6%, 28.0% and 21.9% of the total variation, and all four axes explained 69.1%, 66.8% and 53.6% in June, August and October, respectively. The variation of the species–environment relation explained by the first axis was 40.3%, 36.1% and 35.4% in June, August and October, respectively.

Biplot scalings of CCA with inter-sample distances are shown in figure 5. In June, axis 1 represented a strong gradient caused by DOC (figure 5(a)), indicated by the correlation coefficient of −0.7952. Axis 2 represented a gradient due to the environmental factor nitrite (figure 5(a)), as indicated by its correlation coefficient of −0.5122. Based on the 5% level in a partial Monte Carlo permutation test, DOC (P = 0.0080, F = 3.16) and temperature (P = 0.0240, F = 2.42) were the significant environmental variables, and the two factors alone provided 29.0% and 19.3% of the total CCA explanatory power, respectively. In August, axis 1 and axis 2 represented a gradient caused by nitrite and phosphate (figure 5(b)), indicated by the correlation coefficient of 0.8952 and −0.4834, respectively. Nitrite was the significant environmental variable (P = 0.0140, F = 2.93), which provided 32.0% of the total CCA explanatory power. In October, the variables salinity and pH contributed particularly to the gradient of axis 1 and axis 2 (figure 5(c)), as indicated by the correlation coefficients of −0.8682 and −0.7127, respectively. However, none of the variables displayed significant influence on the bacterial community at the 5% level.

Zoom In Zoom Out Reset image size