Development and microscopic observations of aerobic granular sludge
Activated floccular sludge was used to seed the SBRs, which were operated under conditions optimal for the aerobic granulation process over a period of 11 weeks. The granulation process has five distinct phases: floccular, initiation, maturation, maintenance and dispersal [8]. Here, only three phases of floccular, initiation and maturation phases were observed (Figure 1a).
During Phase I, the floccular biomass had a mean particle size of 51.3 ± 2.2 µm (50th percentile) (Figure 1b). Aerobic granules are typically defined as dense and compact aggregates characterized by a minimum particle size of 100 µm and a SVI5 of 50 mL g-1 or less [24]. Initial decreases in settling time from 120 to 56 min resulted in a 10.5% average loss of biomass (MLSS decreased from 5.0 ± 0.1 to 4.1 ± 0.1 g L-1) by the end of week 1 (Figure 1c). The SVI5 of the floccular sludge increased from 190.8 ± 2.0 to 221.8 ± 5.4 mL g-1, which indicated poor settling of the floccular sludge (Figure 1b).
By Phase II, compact aggregates were observed in the floccular sludge at week 4 and the mean particle size was 96.2 µm (50th percentile) (Figure 1b). Subsequent decreases in settling time from 56 to 24 min did not result in a decrease in overall biomass until week 4 (MLSS increased from 4.9 ± 0.4 to 5.1 ± 0.4 g L-1) when the sludge biomass entered the Phase II. During weeks 4 to 6 of Phase II, the settling time was reduced from 24 to 5 min, which resulted in an average of 23.7% loss of biomass (MLSS decreased from 5.1 ± 0.4 to 3.9 ± 0.5 g L-1) (Figure 1b). This reduction in settling time also coincided with an increase in mean particle size from 108.5 ± 6.9 to 193.0 ± 16.7 µm (50th percentile) (Figure 1b). In addition, the SVI5 also decreased 44% from 112.5 ± 13.2 to 63.0 ± 6.5 mL g-1 (Figure 1c).
By week 7, the sludge biomass had entered Phase III of the aerobic granulation process. The mean particle size of the sludge biomass increased 90% from 193.0 ± 16.7 µm in week 6 to 367.0 ± 68.1 µm in week 7 (50th percentile) (Figure 1b). The particle size and SVI5 of the sludge biomass continued to increase and decrease, respectively, over the remaining weeks. The MLSS of the sludge steadily increased from week 7 onwards (Figure 1c). Over the entire 11 weeks, the reduction in settling time from 120 to 5 min was linked to the appearance of high density and compact sludge particles. This was associated with a mean particle size increase from 51.3 ± 2.2 to 792.4 ± 130.6 µm (Figure 1b). Similarly, the SVI5 decreased significantly from 190.8 ± 2.5 to 16.0 ± 2.1 mL g-1 (Figure 1b). In addition, the MLSS of the sludge also increased from 3.9 ± 0.5 in week 6 to 12.7 ± 0.6 mL g-1 by the end of week 11. These observations indicated that the sludge biomass was mostly in granular form.
Microbial community composition of floccular and granular sludge
Here, the total genomic DNA of the granular sludge was sequenced to track the diversity and changes in bacteria abundance as granulation takes place over 11 weeks of reactor operation. Clustering based on the relative abundance of the microbial communities suggested that in the early floccular stages (weeks 0 and 1), the communities were similar across the 4 SBRs (Figure S1a). However, from week 2, the communities between the reactors diverged, as reflected in changes in the community composition, as the reactors underwent granulation. Despite this, PERMANOVA showed that the reactors are not statistically different from each other (P = 0.184) (Table S1).
The genus ‘Candidatus Accumulibacter’, which is a polyphosphate accumulating organism (PAO) and nitrifier from the phylum Proteobacteria, was the most abundant, with an average increase from 3.6% to 63.53% by week 11 (Figure 2). ‘Candidatus Competibacter’ and ‘Candidatus Contendobacter’, glycogen accumulating organisms (GAOs), did not change appreciably in abundance, between 0.97% to 3.11% and 1.47% to 3.8%, respectively (Figure 2). Nitrifiers, such as Nitrospira, progressively decreased from 16.45% to 6.06% over the course of the experiment. There was a peak of Thauera (a denitrifier) at week 1 at 10.98% but reduced to 3.88% by the end of the experiment. The other members of the top 20 genera generally had a lower abundance with Terrimonas at the lowest between 0.3% and 0.97% (Figure 2).
Bacteriophages exert a complex influence over their microbial hosts and additionally may play a structural component of the matrix [25-27]. Therefore, the relationship between granulation and bacteriophage community dynamics were also investigated here. Only DNA bacteriophages were targeted here and their sequences were assembled into viral contigs to study their relative abundance during granulation (Figure 3a). Microviridae were the most abundant and present in all samples throughout reactor operation at 17% to 99.9%. At the end of the initiation phase (week 7), Podoviridae and Siphoviridae began to significantly increase in abundance and at week 9, were the most abundant viral families after Microviridae at 11.15% ± 1.66% and 8.34% ± 0.96%, respectively. Inoviridae had an increase in abundance to 0.2% ± 0.05% when the sludge developed into compact aggregates (week 5) and peaked at week 9 at 1.46% ± 0.32% (Figure 3a). There was a positive correlation between the increasing granule particle size and the viral counts of Siphoviridae, Microviridae and Myoviridae (Figure 3b). Additionally, a distance based redundancy analysis (dbRDA) was performed to identify covariates which have an effect on the changes in bacterial community using viral family abundance [28]. This analysis suggested the Microviridae and Inoviridae viral families had an effect on the changes in bacterial community composition during the initiation phase (weeks 4 to 7) and maturation phase (weeks 8 to 9), respectively (Figure 3c).
The effect of protozoan predation on aerobic granulation was investigated via total RNA sequencing as metagenomic sequencing did not yield sufficient reads for classification and annotation of eukaryotic sequences beyond the class level (Figure S2). The abundance of the microbial populations was represented by the number of sequencing reads detected per OTU. Mean values were calculated for the number of sequences per OTU to represent the abundance in the four SBRs. A total of 10 OTUs represented approximately 95% of all sequencing reads. Within these 10 OTUs, there were 8 protozoan OTUs which were mostly represented by the genus Telotrochidium (OTU02), class Oligohymenophorea (OTU01, 03 and 04), genus Arcella and order Salpingoecidae (Figure 3d).
The genus Telotrochidium is a group of free swimming peritrichous ciliates while the genus Arcella, and the family of Salpingoecidae represent testate amoebae and flagellates, respectively (Figure 3d). The class Oligohymenophorea represents a large class of ciliated protozoa. Both OTU05 and 08 represented rotifers, which are metazoan predators of suspended microorganisms (Figure 3d). During Phase I, the abundance of Telotrochidium (OTU02) decreased sharply by week 2 and was not detected in most reactors in the following weeks. The family Oligohymenophorea OTU01 also demonstrated gradual decline in abundance from week 0 to 3. Both Oligohymenophorea (OTU 03 and 04) were constantly detected during Phase I in all reactors except in reactor 4, where it was absent at week 03. Salpingoecidae (OTU10), of the flagellate family, was also constantly present from Phase I to III. However, as compact aggregates and granules formed by Phase II and III respectively, Oligohymenophorea (OTU03 and 04) were the most abundant eukaryotic members in the sludge biomass in all reactors. Testate amoeba, including OTU06 and OTU07, were not detected beyond week 5, by which time compact aggregates had formed.
Non-metric multi-dimensional scaling (nMDS) visualization of the eukaryotic communities during granulation demonstrated a high level of dissimilarity between the flocs at week 0 and granules at week 1 (Figure S1c). Based on sludge particle size, the determinant of granulation, the majority of the eukaryotic OTUs, except for Salpingoecidae (OTU10), were positively correlated with the floccular particle size (Figure 3e). In contrast, both Oligohymenophorea (OTU01 and 03) demonstrated a strong positive correlation with the particle size during the initiation and maturation phase. The remaining eukaryotic OTUs had a negative correlation during both the initiation and maturation phases (Figure 3e).
While sequencing provided insights into the eukaryotic communities in the sludge during granulation, microscopic observations were also performed to determine the presence of protozoa and other eukaryotes. Microscopic observations of sludge have also been utilized in membrane bioreactors to compliment sequencing data observations [29]. Swimming ciliates that were most likely Paramecium spp. were observed within the floccular sludge (Figure 4a), while sessile ciliates were attached to the surfaces of the flocs (Figure 4b). These ciliates represent the Oligohymenophorea OTUs detected by sequencing (Figure 3d). Metazoans such as tardigardes (Figure 4c) and rotifers (Figure 4d) were frequently observed in the floccular sludge with crawling ciliates such as Aspidisca sp., circling the Phase I flocs (Figure 4e). These rotifers were likely to be represented by OTU05 and 08 as identified in the sequencing data (Figure 3d). These observations clearly indicated that the inoculum floccular sludge had a diverse community of protozoa present prior to seeding into the SBRs. Upon the formation of compact aggregates at Phase II, no swimming ciliates or large eukaryotes were observed, although rotifers were still occasionally present. Upon granule formation at Phase III, the frequency of crawling ciliates decreased significantly, while sessile ciliates were frequently observed on the granule surfaces (Figure 4f and g). The abundance of sessile ciliates, as determined by microscopy, were also reflected in the sequencing data where there were increases in Oligohymenophorea associated sequences (i.e OTU03 and 04) in most reactors as granules formed.
Development of aerobic granules from untreated and thiram treated floccular sludge
Six mSBRs were seeded with activated floccular sludge and operated under conditions that were optimal for the aerobic granulation process over a period of 8 weeks. To investigate the role of protozoan predation in aerobic granulation, protozoa were removed from the floccular sludge by the addition of 20 mg L-1 thiram to the mSBRs and DMSO was added as a control. The concentration of thiram was previously optimized to minimize any negative effects on the viability of bacteria in the floccular sludge (data not shown). Microscopic observations of control floccular sludge indicated that the conversion of floccular into granular sludge began at week 4 (Figure 5a). Compact aggregates were observed in the initiation phase and these aggregates continued to expand in size. The sludge entered the maturation phase at week 6 and remained in this phase until the end of the experiment at week 8 (Figure 5a). In contrast, thiram treated sludge did not initiate granulation until week 6 and only started to mature by week 8 (Figure 5a).
As the volumes of the mSBRs were too low to allow for particle sizing by the particle size analyser, particle sizing was obtained by quantitative image analysis. The initial mean sludge particle sizes were 84.36 ± 12.41 µm (Figure 5b) and by week 2, the control sludge mean particle size was 89.61 ± 5.94 µm, while the treated sludge was significantly smaller, 67.02 ± 2.65 µm, than the control sludge (Figure 5b). By week 7, the treated sludge was 125.42 ± 10.60 µm, which was similar to the control sludge particles, 122.71 ± 23.00 µm (Figure 5b). By week 8, there was a slight decrease in the control sludge (104.60 ± 17.57 µm), while the thiram treated sludge was significantly larger, 119.36 µm ± 6.05 µm (Figure 5b).
The SVI5 of the treated sludge was significantly higher than the control sludge from weeks 2 to 4 (Figure 5c), suggesting that the thiram treated sludge was less dense and compact and hence required a longer settling time compared to the control sludge. However, from week 5 onwards, the SVI5 for the thiram treated sludge decreased and was not significantly different from the control sludge.
Effects of thiram treatment on microbial communities during aerobic granulation
The microbial communities in the two sludge types were compared by metacommunity sequencing of the V5 region of the 16S and 18S rRNA genes using the Ribotagger method [30]. A total of 30 OTUs, representing approximately 92% of the eukaryotic communities were selected for analysis. Within the inoculum sludge, the eukaryotic communities were dominated mainly by ciliated protozoa OTUs, e.g. OTUs 01, 02 and 03 (Figure 6). As granulation progressed in the control mSBRs, the abundances of these OTUs were consistent, with Oligohymenophorea (OTU01) being the most dominant. Both Oligohymenophorea OTUs 02 and 07 showed a gradual decline in abundance while Oligohymenophorea (OTU26) was not detected beyond week 5. Swimming ciliates from the genus Paramecium (OTU 03) were not detected after week 1 (Figure 6). In contrast to the swimming ciliates, crawling ciliates from the genus Aspidisca (OTU 23) were relatively abundant during granulation. However, these protozoan OTUs were mostly not detected after week 1 in the thiram treated sludge (Figure 6). Interestingly, two flagellate-associated OTUs, OTU08 and 24, increased in abundance in the treated sludge from week 4 onwards.
Metazoan OTUs representing rotifers, e.g. OTUs 04, 05, 06, 15 and 22, were also detected in relatively high abundance in the control sludge and were present throughout the entire granulation process (Figure 6). In contrast, these rotifers were only detected at low abundance during the first three weeks in the treated sludge and were mostly not detected beyond week 4. Eukaryotic communities in the control sludge did not changed drastically over time (Figure S3a). However, eukaryotic communities in the thiram treated sludge diverged over time and were distinctly different from the control sludge from week 1 to 5 (Figure S3a). This was likely due to the absence several dominant protozoan OTUs including OTU01, 02 and 07. Interestingly, the eukaryotic communities in the control and treated sludge began to converge from week 6 onwards, which was likely due to the resurgence of protozoan OTU08 and OTU24 (Figure S3a).
Bacterial OTUs were also analyzed to determine if the absence of predators had any impact on the bacterial communities during aerobic granulation. Based on nMDS visualization, two distinct clusters were observed which indicated dissimilarities between the control and thiram treated bacterial communities during granulation (Figure S3b). The bacterial communities remained relatively similar from week 2 to week 8 in the control sludge while the bacterial communities in the thiram treated sludge continue to change on a weekly basis (Figure S3b). In addition, there was no significant difference in the microbial communities between replicates of the control or thiram sludge due to close clustering in each week (Figure S3b).
In the control sludge, the bacterial communities were dominated mainly by PAOs such as ‘Candidatus Accumulibacter’ (OTU01), GAOs such as ‘Candidatus Competibacter’ (OTU04 and 05) and Nitrospira (OTU06) throughout 8 weeks of aerobic granulation (Figure 7). The abundance of other bacterial members such as Zoogloea (OTU03), Thauera (OTU02), Dechloromonas (OTU07), ‘Candidatus Competibacter’ (OTU12, 17 and 19), ‘Candidatus Contendobacter’ (OTU09), Defluviicoccus (OTU18) and Actinobacteria (OTU20) remained relatively consistent during granulation (Figure 7). In contrast, there was a decrease in the abundance of ‘Candidatus Accumulibacter’ (OTU01) and ‘Candidatus Competibacter’ (OTU04 and 05) from week 1 in the thiram treated sludge. The genus Nitrospira (OTU06) also demonstrated decline in abundance from week 1 onwards with no sign of recovery (Figure 7). The decrease in abundance was also observed in ‘Candidatus Competibacter’ (OTU12, 17 and 19) and ‘Candidatus Contendobacter’ (OTU09) (Figure 7). Interestingly, there were several bacterial OTUs such as Zoogloea (OTU03), Thauera (OTU02), Dechloromonas (OTU07) and Defluviicoccus (OTU18) that increased in abundance from week 1. However, as ‘Candidatus Accumulibacter’ (OTU01) began to gradually increase in abundance from week 5, the abundance of Zoogloea (OTU03), Thauera (OTU02), Dechloromonas (OTU07) decreased. In contrast, OTU18 and 20continued to gradually increase in abundance from week 5 onwards. ‘Candidatus Accumulibacter’ (OTU01) increased in abundance in the thiram treated sludge as it entered Phase II of granulation, where compact aggregates were formed.