Sludge bulkingBulking and foaming are two notorious problems in activated sludge wastewater treatment plants Tboli tribal festivalwhich are mainly associated with the excessive growth of bulking and foaming bacteria BFB. In this study, sludge bulking and foaming sets of high-throughput sludge bulking and foaming HTS of 16S V3—V4 amplicons of 58 monthly activated sludge samples from a municipal WWTP was re-analyzed to investigate the BFB dynamics and further to study the determinative factors. The population of BFB occupied 0. Pair-wise correlation analysis and canonical correlation analysis CCA showed that Gordonia sp. Bacteria species correlated with BFB could be clustered into two negatively sudge modules.
Activated Sludge: Bulking and Foaming Control - Jiri Wanner - Google Books
Bulking and foaming are two notorious problems in activated sludge wastewater treatment plants WWTPs , which are mainly associated with the excessive growth of bulking and foaming bacteria BFB. In this study, data sets of high-throughput sequencing HTS of 16S V3—V4 amplicons of 58 monthly activated sludge samples from a municipal WWTP was re-analyzed to investigate the BFB dynamics and further to study the determinative factors.
The population of BFB occupied 0. Pair-wise correlation analysis and canonical correlation analysis CCA showed that Gordonia sp. Bacteria species correlated with BFB could be clustered into two negatively related modules. Moreover, with intensive time series sampling, the dominant BFB could be accurately modeled with environmental interaction network, i.
Bulking affects the settleability of bioflocs, which may result in failure of solid-liquid separation 3 while foaming on the water surface of aeration tank needs extra operation, lowers the effluent quality and causes loss of biomass 4. Even though these issues have experienced extensive amount of research by both improving configuration of process and controlling the relevant filamentous bacteria, there is still no systematic method to treat them and they still occur sporadically all over the world 5 , 6 , 7.
The bulking and foaming bacteria BFB are deemed as those bacteria overgrowth in a sludge bulking or foaming episode. Their roles in sludge bulking or foaming are not well studied, although their physical roles in the floc formation are well documented as the backbone of flocs in AS 7 , 8 , 9. In recent years, large-scale microbial communities profiling with high throughput sequencing HTS of 16S rRNA amplicons 14 , 15 and whole environmental DNA 16 , 17 have become powerful tools to investigate microbes in environmental samples.
Pair-wise correlation Mostly Pearson or Spearman correlation based network analyses among microbial communities and environmental parameters by HTS of 16S rRNA markers in various environmental niches such as soil, human gut and AS have been conducted to reveal environment-microbe and microbe-microbe associations, which have deepen our understanding of the determinative factors and the taxonomic relatedness on microbial communities 18 , 19 , This method could explain the possible linear Person or rank linear Spearman relationships between microbial communities and environmental parameters.
For nonlinear relationships, the environmental interaction network EIN method had been proven as an effect way to model the dynamics of microbial assemblages in Western English Channel ocean region 21 , which is different from other modeling methods in WWTPs which focused on prediction of effluent quality and sludge volume index SVI with physicochemical parameters and operational parameters.
To the best of our knowledge, currently there are no studies using large-scale HTS of 16S rRNA marker time series data to specifically investigate the associations between abiotic environmental parameters , biotic other related bacteria factors and BFB. In this study, 16S V3—V4 pyrosequencing data sets 22 of 58 samples collected over five years from to were re-analyzed for BFB profiles. Over the whole sampling period, the WWTP was efficient in CBOD removal however suffered by unstable ammonium removal and periodically foaming in winter each year As shown in Fig.
Actinobacteria , CFB sp. The most abundant foaming bacteria was Gordonia sp. The most abundant bulking bacteria was Nostocoida. The four most abundant BFB were all from phylum Actinobacteria. Apart from the dominant four types, other BFB were presented sporadically across all the time with relatively low abundance. Heatmap values were transformed to log2. Bulking and foaming bacteria were grouped with phyla. In a survey of Italy WWTPs, Microthrix pavicella was the dominant filamentous microorganism involved in sludge bulking or foaming 23 ; however, in the present study, it was Gordonia sp.
The frequency distribution graph Fig. S1 of the dominant six BFB showed quite different frequency patterns. The most abundant species Gordonia sp. The abundance pattern was different from the results of a previous survey of Danish WWTPs using quantitative FISH which showed that only minor changes in relative abundance over three years 7.
The abundance of BFB in Shatin over five years had bigger variations compared with that in our previous studies of 14 sewages plants 15 which were from 1. The discrepancy indicated that the abundance variations of the bacteria related with bulking and foaming in Shatin WWTP over time which suffered from periodical sludge bulking and foaming in winter and spring could be larger than that from geographically distributed WWTPs from different countries.
The complexity of different frequency patterns for different BFB revealed the generalist and specificity of BFB along time. However, these limitations could be alleviated with the increase of sequencing length or application of third generation sequencing platform which generates sequences longer than full length of 16S rRNA Thus, the identification of BFB using 16S profiles will be further applied in the future studies.
Overall, total BFB showed seasonal variations with higher abundance in winter-spring and relatively lower abundance in adjacent summer-autumn except for over the five years, which was mainly contributed by the variation of Gordonia sp. Post-hoc Tukey HSD test was performed for each pair of seasons. For each BFB, in details, Gordonia sp. An annual decline trend of the portion of total BFB in winter was also observed Fig.
Although bacteria Gordonia amarae has been reported largely associated with sludge foaming 27 , in the present study, the existence of Gordonia sp. The discrepancies between these two studies could be due to influent characteristics, climate conditions of the two regions, and operational parameters.
To investigate the influence of operational parameters on the dynamics of BFB, Pearson and Spearman coefficient indexes were calculated between the most abundant four BFB and operational parameters.
In the present study, several strong correlations have been identified by correlation analysis, including Gordonia sp. These correlations have never been reported before. The above results were confirmed by Canonical corresponding analysis CCA results, which showed that Gordonia sp.
There were reports showed new strains in genus Gordonia could be nitrate reducing bacteria 28 , 29 , and this indicated that the Gordonia sp. We analyzed the draft genome of the novel Gordonia sp. The negative correlation between Tetrasphaera sp.
The dependence of foaming bacteria Microthrix parvicella on temperature had been observed in previous study 30 which was consistent with our result. Due to the limitation of technologies used before, it was difficult to conduct large-scale correlation analysis of filamentous bacteria with operational parameters to identify the sensitive parameters. In the present study, we obtained several strong correlations of the operational parameters with specific BFB by the statistical analysis based on long-term time series profiling of the bacterial community in the full-scale WWTP, although these novel relationships need to be further confirmed by experiments.
We also preliminarily demonstrated that the correlation of Gordonia sp. Apart from the abiotic factors, biotic interactions were also quite important for the dynamics of bacteria assemblages 21 , Inside each module, bacteria species were all positively correlated; between the two modules, they are negatively correlated. The two modules both were composed with widely distributed phyla.
Detailed analysis of the two modules showed that Specifically, seven species from order Rhizobiales were all strongly and negatively correlated with Gordonia sp. The species directly correlated with Tetrashpaera sp. Nodes were bacteria and edges were the correlation between bacteria.
Green edges represented positive correlation and red edges were negative correlation. All the correlated bacteria were clustered into two co-exclusive modules. Generally, in microbe-microbe interaction, negative correlation may be caused by prey-predator, competition, amensalism, different preference in living environment and so on Although it is not an easy task to find the reasons for these correlation patterns in the network, the non-random positive and negative correlation of different bacterial species with BFB pose novel knowledge to BFB related associations in AS system.
Since correlation based analysis can only capture potential linear or rank linear relationships between BFB and the related abiotic and biotic factors, to explore possible non-linear relationships of them, we constructed EIN which was a Bayesian network generated with environmental parameters, interactions between BFB and other bacteria. The Bayesian network was probabilistic graphical model which represented conditional dependence relationships among a group of random variables.
Thus, edges in the network possibly referred to causal-relationships between the parent node and the children node. Results of the derived functions for the three selected BFB and their accuracies were listed in Table S2. Figure 5a showed the predicted and observed Gordonia sp. The figure showed that the prediction accuracy of EIN derived function was much better compared with the function derived only by the environmental parameters.
Prediction accuracy was fairly good with an R 2 of 0. S3 , Table S2. It should be noticed that the related bacteria with BFB in the EIN module were not exactly identical to those identified using the Spearman correlation based network.
This was mainly due to the ability of the Bayesian inference in identifying nonlinear relationship among variables. Overall, the incorporation of biotic interaction in the EIN model had better accuracy than that one with only environmental parameters indicating that the biotic factors were also important factors in determination of the population of BFB.
In conclusion, pyrosequencing of 16S rRNA gene revealed high diversity 17 types of bulking and foaming bacteria occurring in a full scale WWTP in five years, occupying 0. Total BFB showed significant seasonal variations with higher abundance in winter than summer and the variation was mainly contributed by Gordonia sp.
CCA showed consistent results with the above correlation analysis. Bacteria correlated with BFB could be clustered into two modules; the two modules were negatively correlated with each other and positively correlated inside each module. Correlations between Gordonia sp. The AS was 1: Wastewater characteristics and operational parameters were collected accordingly from Drainage Services Department. Then ChimeraSlayer algorithm 34 was used to identify chimera sequences.
After removing chimera sequences, the clean sequences of each sample were normalized to 6, To investigate bacterial species correlated with BFB, we firstly remove all the BFB sequences and then the remaining sequences were clustered into operational taxonomy units OTUs at 0. Canonical corresponding analysis CCA was generated by Canoco4. An EIN was a Bayesian network BN with both environmental parameters and microbial interactions as proposed in a study using EIN to predict the microbial community of ocean with time series data Due to different units for environmental parameters, all the environmental parameters were transformed to 1 to by the following equation for normalization After filtering, only the most abundant three BFB met the requirement.
For the detail running parameters of BANJO, a maximum of five parents, All Local Moves proposer, simulated annealing and randomly configured networks were used.
The EIN we finally obtained was a directed acyclical graph DAG whose edges represented causal relationship between the parent nodes and their child nodes inferred by the observed data. The relationship in the EIN could be seen as an artificial neural network. Then each BFB can be expressed as a function of its parent nodes.
The function was deduced using Eureqa 0. How to cite this article: The authors would like to thank Dr. National Center for Biotechnology Information , U. Published online Apr Find articles by Xiao-Tao Jiang. Find articles by Feng Guo.