Bacterial community composition. In order to study the microbial structure of the biofilm and suspended biomass that were developing in the IFAS-MBSBBR reactor, a total of 15 samples were taken at intervals from an experiment lasting 573 days. The microbiome of both environments was described at the phylum and genus level. A total of 26 bacterial phyla and 783 bacterial genera were identified. The most numerous phyla and genera in the biofim and suspended biomass samples are presented in in Figures 1 and 2. Both in the biofilm and the suspended biomass, the most numerous phyla were Proteobacteria, with respective mean abundances of 39.3% ± 9.0 and 40.8% ± 8.2, and Bacteroidota, with respective mean abundances of 14.2% ± 4.9 and 26.1% ± 13.7. Additionally, the phylum Chloroflexi was rather abundant in the biofilm (with a mean abundance of 13.9 ± 8.1), while Actinobacteriota and Patescibacteria were relatively abundant in the suspended biomass (with mean abundances of 9.0% ± 9.6 and 7.5% ± 8.1 respectively). STAMP analysis identified significant overrepresentations of Chloroflexi, Acidobacteriota and Nitrospirota in biofilm and of Firmicutes in suspended biomass.
In both environments, the abundance of various groups of bacteria changed over time. In the biofilm, the abundance of Proteobacteria and Actinobacteria gradually decreased, while that of Chloroflexi increased. In the suspended biomass, the changes in abundance were larger and more rapid, and the abundance of Bacteroidota changed to the largest extent, ranging from 12.7% after 42 days of reactor operation to 52.3% after 110 days, when it was the predominant phylum. The abundance of Patescibacteria also changed substantially: its abundance was highest on 78th, 205th and 447th day of the process, reaching values of 20.1%, 11.0% and 7.2%, respectively. Similar changes took place in the abundance of Armatimonadota, which reached 11.4% and 7.6% on 547th and 573th day, but did not exceed 0.1% in the samples taken at other times.
At the genus level, the less abundant genera (each <1.5% of the total bacterial community) combined to constitute the largest shares in all samples of biofilm and suspended biomass samples (mean abundance of 45.6% ± 5.8 and 30.5% ± 6.0, respectively). Initially, Ornithinibacter was relatively abundant in the biofilm, which is the reason for its fairly large mean abundance of 4.3% ± 5.3%. Over time, however, the abundance of this group decreased substantially, and at the end of the process, it was only 0.3%. Similarly, the abundance of Rhizorhapis was 13.92% in the first sample, but then it decreased and this genus was not detected after 205th day. The changes in the abundance of Nitrospira and Candidatus Competibacter are also noteworthy, increasing at first and then decreasing. Nitrospira was most abundant in the sample from 25.02.20 (5.7%), and Candidatus Competibacter, in the sample from 26.03.19 (6.4%). The abundance of the remaining genera did not exceed 5% at any time during this study.
In the samples of suspended biomass, the abundance of Ornithinibacter also decreased significantly at the beginning of the experiment (from 23.0% in the first sample and 12.5% on the 78th day to values below 3% in subsequent periods). Generally, the abundances of individual genera changed more rapidly in the suspended biomass than in the biofilm. There were also rapid decreases and increases in the abundance of many groups of bacteria in the following periods, particularly in the case of uncultured Saccharimonadales and Zoogloea. Figure 3 shows groups of bacteria that differed significantly between biofilm and sludge samples. Denitratisoma, Nitrospira, Candidatus Competibacter, Dechlorosoma, Candidatus Accumulibactrer and Kouleothrix were significantly more abundant in the biofilm than in the biomass, while Zooglea, uncultured Saccharimonadales, Rhodobacter and Ottowia were significantly less abundant in the biofilm.
Bacterial diversity. Bacterial community indices were estimated using the EZBioCloud platform (Table 1). The average Good’s coverage of all samples was 99.75% ± 0.047%, indicating that the sequencing coverage was very high. The total number of OTUs differed between samples and types of biomass. The mean number of OTUs for biofilm was 1614 ± 141 and 993 ±109 for suspended biomass. The Chao1 index was used to evaluate community richness, i.e., the number of species in the biofilm and suspended biomass communities, and the Shannon index was used to measure community diversity, taking into account both the abundance and the evenness of the species. The mean values of these indices indicated that the biofilm community was richer and more diverse than the suspended biomass community (Chao1: 1734.64 ± 138.59 vs. 1105.72 ± 138.59; Shannon: 5.34 ± 0.23 vs. 4.27 ± 0.41). The differences between communities were all statistically significant (P<0.05).
Figure 4 shows the results of beta diversity analysis based on the Bray-Curtis dissimilarity. Principal Coordinates Analysis showed that the biofilm and suspended biomass samples grouped into two separate clusters, although the distances between individual samples were quite large. Hierarchical analysis indicated the development of biofilm and suspended biomass was independent and confirmed the distant differences between these two types of biomass.
To model interactions and relationships between bacteria in the biofilm and suspended biomass, co-occurrence network analysis was used. Networks are graphical models that represent bacterial communities. The nodes of the network symbolize a taxonomic group. Nodes in networks are connected by edges, when a statistically significant relationship exists between them. In the present study, two networks were created that represent
the co-occurrence of genera in the biofilm and suspended biomass. In Figures 5 and 6, the color of each node is based on its modularity class parameter, and its size is based on its betweenness centrality. The basic parameters characterizing both networks are presented in Table S1. In general, the biofilm network had more connections between nodes than the suspended biomass network, and the distance between nodes was smaller in the biofilm network, indicating that the microorganisms creating the biofilm are more closely related and have more relationships between them. Both networks had the same number of nodes (83), but the biofilm network had more edges (connectors between nodes symbolizing
co-occurrence). In both networks, the number of positive associations was slightly higher than that of negative associations, accounting for 55% of the total number of connections. The mean clustering coefficient (i.e., the ratio between the observed and the maximum possible number of links between a node and its neighbors) was higher for the biofilm than for the suspended biomass (0.556 vs. 0.432). Similarly, the network density, which is the ratio between the observed number edges and the maximum possible number of them, was higher for the biofilm (0.073 vs. 0.05). The network diameter (the distance between the two most distant nodes) was shorter for the biofilm than for the biomass (6 vs. 7). Likewise, the average path length, which is the number of edges in the shortest path between pair of nodes, was shorter in biofilm network than in suspended biomass network (1.984 vs. 2.241). The mean node degree (the number of edges between one node and other nodes in the network) was greater in the biofilm network than in the suspended biomass network (6.012 vs. 4.12). Node degree ranged from 1 to 31 in the biofilm network and from 1 to 23 in suspended biomass network. In the biofilm network there were four nodes with the highest degrees (≥ 30) that can be considered hub nodes: Diaphorobacter, Rhizorapis, Mesorhizobium and Pseudoxanthomonas. These microorganisms had 61.5% positive and 38.5% negative connections with other microorganisms. Interestingly, although the abundance of Mesorhizobium and Pseudoxanthomonas was low (not exceeding 1.5% of the total bacterial community in any sample) they had positive associations with highly abundant bacteria, e.g., Ornithinibacter. The suspended biomass network also had 4 hub nodes (with node degree ≥ 20): Nocardioides, Gemmatimonas, Leptothrix and Rhizorhapis. These hub nodes were connected to other nodes by similar amounts of positive and negative edges (51.8% and 48.2%, respectively). The size of the nodes in created networks is proportional to their betweenness centrality (a parameter that indicates the frequency of occurrence of a particular node on the paths between two other nodes). High values of betweenness centrality indicate that a node has a central location in a network, while low values indicate that it has a peripheral location[9]. Microorganisms with high betweenness centrality play key roles and act like bridges between other bacteria in the network. In the biofilm network, Paracoccus, Phaeodactylibacter and Pseudoxanthomonas had the highest values of betweenness centrality, whereas in the suspended biomass network, Dongia, Diaphorobacter and Rhizorhapis had the highest values.
The networks were constructed with additional nodes representing the efficiency of pollutant removal processes, i.e., removal of organic and phosphorus compounds, as well as denitrification, ammonification and nitrification. In the biofilm network, the efficiencies of phosphorus compound removal and of nitrification had the most associations with microbial nodes (9 positive and 2 negative, and 6 positive and 4 negative, respectively). The efficiency of phosphorus removal was positively associated with the abundance of Candidatus Accumulibacter, Dechlorosoma, Thauera and uncultured Saccharimonadales, while nitrification was positively associated with the abundance of taxa such as Nitrosomonas, Sphingomonas and Thermomonas. The remaining efficiencies of pollutant removal had no or only 1-2 connections with microbial nodes. In the suspended biomass network, in contrast, all efficiency nodes were connected with those of microbes, with degree ranging from 2 to 9. The nitrification and ammonification efficiency nodes had the highest degree and were positively associated with, for example, Nocardioides, Rhodobacter and Sphingomonas. Organic compound removal efficiency was positively associated with Blastocatella, Ornithinibacter and Terrimonas, whereas in the biofilm network, it had no edges.