Seasonal Dynamics of the Activated Sludge Microbiome in Sequencing Batch Reactors, Assessed Using 16S rRNA Transcript Amplicon Sequencing

Sequencing batch reactors are a common design for wastewater treatment plants, particularly in smaller municipalities, due to their low footprint and ease of operations. However, like for most treatment plants in temperate/continental climates, the microbial community involved in water treatment is highly seasonal and its biological processes can be sensitive to cold temperatures. The seasonality of these microbial communities has been explored primarily in conventional treatment plants and not in sequencing batch reactors. Furthermore, most studies often only address which organisms are present. However, the activated sludge microbial community is very diverse, and it is often hard to discern which organisms are active and which organisms are simply present. In this study, we applied additional sequencing techniques to also address the issues of which organisms are active and which organisms are growing. By addressing these issues, we gained new insights into seasonal microbial populations dynamics and activity patterns affecting wastewater treatment.

ABSTRACT

Activated sludge is comprised of diverse microorganisms which remediate wastewater. Previous research has characterized activated sludge using 16S rRNA gene amplicon sequencing, which can help to address questions on the relative abundance of microorganisms. In this study, we used 16S rRNA transcript sequencing in order to characterize “active” populations (via protein synthesis potential) and gain a deeper understanding of microbial activity patterns within activated sludge. Seasonal abundances of individual populations in activated sludge change over time, yet a persistent group of core microorganisms remains throughout the year which are traditionally classified on presence or absence without monitoring of their activity or growth. The goal of this study was to further our understanding of how the activated sludge microbiome changes between seasons with respect to population abundance, activity, and growth. Triplicate sequencing batch reactors were sampled at 10-min intervals throughout reaction cycles during all four seasons. We quantified the gene and transcript copy numbers of 16S rRNA amplicons using real-time PCR and sequenced the products to reveal community abundance and activity changes. We identified 108 operational taxonomic units (OTUs) with stable abundance, activity, and growth throughout the year. Nonproliferating OTUs were commonly human health related, while OTUs that showed seasonal abundance changes have previously been identified as being associated with floc formation and bulking. We observed significant differences in 16S rRNA transcript copy numbers, particularly at lower temperatures in winter and spring. The study provides an analysis of the seasonal dynamics of microbial activity variations in activated sludge based on quantifying and sequencing 16S rRNA transcripts.

IMPORTANCE Sequencing batch reactors are a common design for wastewater treatment plants, particularly in smaller municipalities, due to their low footprint and ease of operations. However, like for most treatment plants in temperate/continental climates, the microbial community involved in water treatment is highly seasonal and its biological processes can be sensitive to cold temperatures. The seasonality of these microbial communities has been explored primarily in conventional treatment plants and not in sequencing batch reactors. Furthermore, most studies often only address which organisms are present. However, the activated sludge microbial community is very diverse, and it is often hard to discern which organisms are active and which organisms are simply present. In this study, we applied additional sequencing techniques to also address the issues of which organisms are active and which organisms are growing. By addressing these issues, we gained new insights into seasonal microbial populations dynamics and activity patterns affecting wastewater treatment.

INTRODUCTION

The activated sludge process is a biological nutrient removal process in wastewater treatment plants (WWTPs) relying on taxonomically and metabolically diverse microbial consortia of which the majority of microorganisms have yet to be cultured and carry out unknown functions (1). Understanding seasonal dynamics and activity of individual microbes in this complex community is paramount to future sustainable nutrient management and economic wastewater treatment (2). Activated sludge is dependent on specific core microorganisms to obtain desired process outcomes. The activated sludge microbiome has been extensively studied (3–13) and facilitated development of an ecosystem-specific reference database (14). However, activated sludge is dependent on physiological activity of specific microbial functional groups; hence, it is essential to combine information on microbiome taxonomy and classification with phenotypic information on the physiological state of identified microbial populations.

16S rRNA gene amplicon sequencing is the standard method for identification and characterization of the diversity, composition, and dynamics of microbial communities. However, most sequencing studies are based on amplifying rRNA genes from environmental DNA extracts. Environmental DNA extracts contain DNA from active cells, inactive but viable cells, and dormant cells, as well as extracellular DNA from degraded or lysed cells. Therefore, analyses based on environmental DNA presents a one-sided view of the community structure lacking information on the metabolic state of microbial populations (15). In contrast environmental RNA provides a more immediate view on microbial activities because RNA has a rapid turnover rate within cells (16) and a shorter half-life than DNA (17). Thus, sequencing reverse transcripts (cDNA) of 16S rRNA is more representative of the active fraction of the total microbial community (18). The number of sequencing tags of a specific taxon in an RNA-based 16S rRNA library relative to a DNA-based 16S rRNA gene library can be used as a measure of potential activity. Therefore, parallel sequencing and comparison of 16S rRNA amplicon libraries based on RNA and DNA can help to (i) reveal shifts in activity patterns among rare and abundant taxa, (ii) identify populations in the activated sludge community changing their physiological state during different SBR cycles, and (iii) allow new insights into seasonal variations in physiological states of core members of the activated sludge community.

This approach has been used in previous studies of microbiomes in soils and lakes and provided the basis for generating and testing hypotheses on the complex interplay of microbial community structure and activity (19–24). While rRNA as measure of microbial activity has limitations (18), rRNA is a critical component required for protein synthesis. The ratio of 16S rRNA transcripts and 16S rRNA gene amplicons of a specific taxon can therefore serve as a proxy of potential for de novo protein synthesis.

Previous studies have reported the activated sludge core community (4–11) mostly based on the abundance and occurrence frequency of operational taxonomic units (OTUs). Griffin and Wells reported on a core microbial community of 134 OTUs across six wastewater treatment plants in the Midwest of the United States. The core community they observed was also highly synchronous by season (7). Saunders et al. analyzed the microbial community composition of 13 wastewater treatment plants in Denmark. They identified 63 core genus-level OTUs and found numerous OTUs in the influent wastewater which seemed to continuously seed the activated sludge microbial community (10). Interestingly, Saunders et al. reported that Nitrotoga was the dominant nitrite-oxidizing bacterial taxon in their analyzed systems, while many other published studies on the activated sludge microbial community reported Nitrospira as a ubiquitous taxon of nitrite-oxidizing bacteria across global wastewater treatment plants (4). The Global Water Microbial Consortium noted 28 core OTUs which were prevalent in most wastewater treatment plants across the world and considered core OTUs to be prevalent in more than 80% of the wastewater treatment facilities worldwide. While speculations on the existence of a global core activated sludge microbiome are intriguing, the findings also raised questions as to why there are consistent deviations from the global activated sludge microbiome consensus (3, 4, 10) and what causes these alterations in microbial community assembly.

For the WWTP under investigation in this study, we have also previously reported on the core community, as determined by sequencing and comparing the occurrence and abundance of 16S rRNA gene amplicons from three parallel sequencing batch reactors (SBRs) sampled weekly for more than a year (3). We identified core OTUs which correlated with both seasonal temperature variations and nitrification performance. While ammonia removal drastically declined in cold temperatures, we found that this seasonal variation in plant performance was not correlated with a decrease in the abundance of OTUs capable of ammonia oxidation. We reported 114 OTUs which were consistently present in the SBRs and classified these OTUs as the core microbiome. While all these previous studies have expanded our understanding of the diversity of activated sludge, many of the identified taxa are not known or have not been characterized with respect to metabolic function and physiological role in wastewater treatment. This issue is also discussed in the Microbial Database for Activated Sludge (MiDAS) field guide, which focuses on the curation of microorganisms performing important functions in WWTPs (14). We therefore decided to evaluate if 16S rRNA transcript sequencing alongside routine DNA-based 16S rRNA gene amplicon sequencing and absolute quantification of microbial biomass by quantitative PCR can inform and guide the identification of process-relevant key players in the diverse activated sludge core microbiome. Here, we report how this approach helped us to identify core OTUs which are not only present but also active and proliferating in the dynamic environment of SBRs.

In order to expand the core microbiome definition with these new criteria, we set out to analyze how the activated sludge microbiome fluctuates across seasons and reactor cycles. Seasonal data have the potential to reveal compositional changes influenced by plant operational and environmental parameters, such as seasonal temperature changes. Using full-scale SBRs enables observations across hydraulic residence times, which are observable only in batch and plug flow reactors and not in continuously stirred tank reactors (25). Throughout seasons and reactor cycles, we sequenced the 16S rRNA gene as a marker gene for composition and 16S rRNA transcripts, which serve as a proxy for the protein synthesis potential (18). Additionally, the 16S rRNA transcript copy numbers were quantified using reverse transcription-quantitative PCR (RT-qPCR) at different reactor cycles to discern overall variations in protein synthesis potential. Our data provide insight into the activity and composition of the activated sludge community between seasons and help to identify seasonal variations in activity profiles among taxa. This report presents a comprehensive analysis of the activated sludge microbiome comparing three parallel full-scale SBRs. We describe activated sludge community shifts comparing inflow and different reactor cycles based on the 16S rRNA gene and 16S rRNA transcript amplicon sequencing. Further, we performed our analysis on a WWTP in a continental climate which undergoes strong seasonal temperature variations (3). This allowed us to compare samples collected during the four seasons with respect to community shifts and plant performance parameters. By combining DNA and RNA amplicon sequencing and RT-qPCR, we gained insights into the assembly and seasonal shifts of the activated sludge community. We outline new criteria for analyzing activated sludge core microbiomes, comprising presence, activity, and growth. Our approach can further inform researchers on the importance and role of many yet-to-be-cultured and uncharacterized taxa in activated sludge.

RESULTS

Plant performance.

Data records provided by Brainerd Wastewater Treatment Facility were used to tabulate average plant operational parameters and performance metrics during a 30-day period around each of the four sampling dates (Table 1). Temperatures and volumetric flow rates were measured daily (n = 30 per month), whereas other parameters were measured three times per week (n = 12 per month). Both influent water temperature and air temperature were statically different across seasons. The influent and effluent wastewater volume in fall was significantly higher than in other seasons, likely due to tourism in the area. The chemical composition of the influent wastewater remained highly similar and did not significantly change for most parameters over the seasons. Exceptions were the effluent ammonia and total suspended-solid concentrations. Total suspended solids appeared to have higher concentrations in the summer and fall. However, these data points seemed to be largely driven by outlier data points of very high concentrations measured just once per month (612 mg/liter in summer and 1,073 mg/liter in fall). We decided to not remove these data points since they are public records recorded by the wastewater operators. Suspended-solid concentrations peaked always after our sampling days and were unlikely to impact our results. Effluent ammonia was measured only once per month, and therefore, statistical inferences could not be made. We have previously shown detailed annual changes in effluent ammonia, which is known to be significantly impacted during cold temperatures in what is referred to as seasonal nitrification failure (3).

TABLE 1.

Operational conditions and performance metrics for the sample trips taken throughout 2017 and 2018^a

Parameter	Value for season (sample date)
	Summer 2017 (1 July 2017)	Fall 2017 (3 October 2017)	Winter 2017 (27 December 2017)	Spring 2018 (27 March 2018)
Inf. water temp (°C)	17.53 ± 0.51	15.93 ± 0.91	13.89 ± 0.90	11.56 ± 0.58
Air temp range (°C)	13.3 to 27.8	11.7 to 16.1	−33.9 to −20.6	−18.8 to −8.1
Inf. (m3/day)	9,774 ± 318	12,911 ± 1,591	9,956 ± 409	9,410 ± 364
Eff. (m3/day)	8,910 ± 455	12,547 ± 1,818	9,138 ± 591	8,638 ± 409
Inf. BOD5 (mg/liter)	175 ± 52	167 ± 64	139 ± 48	164 ± 29
Eff. BOD5 (mg/liter)	2.00 ± 0.00	1.23 ± 0.60	1.33 ± 0.49	1.75 ± 0.27
Inf. TSS (mg/liter)	338 ± 138	378 ± 261	169 ± 52	263 ± 47
Eff. TSS (mg/liter)	2.83 ± 1.34	2.85 ± 2.41	1.92 ± 0.67	2.92 ± 1.38
Inf. Phos (mg/liter)	4.87 ± 0.44	3.58 ± 0.67	5.08 ± 0.30	5.51 ± 0.47
Eff. Phos (mg/liter)	0.20 ± 0.06	0.33 ± 0.23	0.15 ± 0.06	0.18 ± 0.02
Eff. NH3 (mg/liter)b	0.10	0.10	4.12	9.43
MLSS (mg/liter)	1,378 ± 166	1,335 ± 152	1,165 ± 219	1,092 ± 167

^aThe data are averages from a 30-day window around the sample period. The influent (Inf.) water and air temperatures, and flow rates, had periods of statistically significant differences compared to other seasons. Other than effluent (Eff.) ammonia, typical effluent operational performance metrics for biological oxygen demand, phosphorus (Phos), and total suspended solids did not statistically vary throughout the seasons. BOD5, biochemical oxygen demand; TSS, total suspended solids; MLSS, mixed liquor suspended solids.

^bValues were recorded only once per month. The date closest to the sampling trip was used.

Quantification of 16S rRNA transcript and 16S rRNA gene abundance during reactor cycles.

We quantified 16S rRNA transcript abundances during the reactor cycle at 10-min intervals at all four seasons (Fig. 1). The average 16S rRNA transcript abundances (log transcripts per milliliter) during the static filling (SF) were 11.4 ± 0.5 in summer, 13.4 ± 1 in fall, 11.7 ± 1 in winter, and 11.9 ± 0.5 in spring. When the influent wastewater and activated sludge were mixed during mixed fill (MF), the 16S rRNA transcript abundances reached 12.3 ± 0.7 in summer, 13.8 ± 0.3 in fall, 12.5 ± 0.9 in winter, and 12.5 ± 0.4 in spring. The mixing led to statistically significant increases in 16S rRNA transcript abundances in summer, winter, and spring (all P < 0.05), while transcript abundance in fall remained constant (P = 0.12). During the 2-h-long aerobic react cycle (with phases R1 and R2) the average 16S rRNA transcript abundances were 11.9 ± 0.9 in summer, 13.7 ± 0.5 in fall, 11.5 ± 0.9 in winter, and 12.9 ± 0.4 in spring. At the onset of aeration, the samples collected in winter experienced a statistically significant decline in 16S rRNA transcript abundance (P < 4.5e−5). During the aerobic react cycle, 16S rRNA transcript abundances in the summer were significantly lower than in spring (P < 1.8e−5) but higher than in the winter (P < 0.025). During settling, the bottom sludge (SS) averaged 16S rRNA transcript abundances of 13.2 ± 0.5 in summer, 14.5 ± 0.4 in fall, 13.1 ± 1.0 in winter, and 12.9 ± 0.3 in spring. In the supernatant water, we quantified 16S rRNA transcript abundances of 11.1 ± 1.1 in summer, 12.3 ± 0.8 in fall, 11.3 ± 0.8 in winter, and 10.9 ± 0.8 in spring.

FIG 1.

The activity of the influent wastewater and activated sludge community is shown for each season and throughout the reactor cycle in 10-min intervals. Panel a shows the SF, MF, R1, and R2 cycles, while panel b and c time points occur simultaneously, as the activated sludge settles (SS) to the bottom (b) and decants (DW) from the top (c) of the reactor. The quantitative abundance of 16S rRNA transcript (solid lines) is shown by season (summer in red squares, fall in orange circles, winter blue triangles, and spring in green diamonds). The solid hatch represents the standard deviation for the 16S rRNA transcript. The dashed line and dashed gray hatch represent an average abundance of 16S rRNA genes throughout the year, which does not significantly change with respect to seasons. The transcript abundances in the fall are continuously significantly higher transcript copy numbers for 16S rRNA, while the other seasons remain relatively comparable at different stages. Each season experiences a significant increase in transcripts once the activated sludge is mixed with the influent wastewater, followed typically by a period of stabilization throughout the react cycles. Winter experienced a significant decrease in 16S rRNA transcripts at the onset of aeration. Settling causes an increased concentration of activity in the thick bottom settled sludge, while the decant significantly declines before moving on in the treatment process.

The 16S rRNA gene abundance did not show any statistically significant seasonal differences when different reactor cycles were compared. The average 16S rRNA gene abundances ranged from 8.6 ± 0.2 log genes/ml during SF to an average of 9.6 ± 0.42 log genes/ml during MF, R1, and R2.

Beta diversity analysis.

In Fig. 2a, PC1 (x axis) represents 38.2% of the total variance explained and separates the inflow community (SF) from activated sludge samples. PC2 (y axis) represents 17.3% of the total variance explained and separates 16S rRNA gene sequences and 16S rRNA transcripts. Within PC2, we also observed separation of the seasons within the activated sludge compositions. The 16S rRNA gene and 16S rRNA transcript compositions were most different in fall and spring. The activated sludge 16S rRNA transcript community composition in winter and summer overlapped; however, their seasonal 16S rRNA gene community compositions did not.

FIG 2.

Principal-coordinate analysis was used to compare trends in reactor cycles and community composition and expression, as well as seasonal trends. Panel a highlights the difference between the influent wastewater and the activated sludge community. Panel b removes the static fill to represent the distinct separation of 16S rRNA genes and 16S rRNA transcripts in the activated sludge over the x axis, while the seasons are separated on the y axis. The abbreviations correlate to the six reactor cycles sequenced in the study.

We removed SF data and recalculated dissimilarity to emphasize differences within the activated sludge composition throughout reactor cycles (Fig. 2b). PC1 (x axis) represents 32.2% of the total variance explained and separates the 16S rRNA gene communities from the 16S rRNA transcripts. PC2 (y axis) represents 18.2% of the total variance explained and shows separation of communities by season. The activated sludge 16S rRNA transcript compositions from winter and summer overlapped; however, the 16S rRNA gene community composition did not. Additional dissimilarity analysis was performed to separate the 16S rRNA genes from the transcripts in order to determine if further trends in separation across specific reactor cycles would become apparent (see Fig. S5 in the supplemental material). In Fig. S5a, PC2 (y axis) showed a slight separation of the decanted water (DW) samples from the other reactor cycle’s 16S rRNA gene compositions.

Seasonal shifts of 16S rRNA gene and transcript community composition.

Figure 3 shows the log ratio of the relative abundance of individual OTUs between seasons based on 16S rRNA genes (x axis) and 16S rRNA transcripts (y axis). OTU distribution along the x axis denotes log fold change in relative abundance of 16S rRNA genes between seasons. Distribution along the y axis indicates log fold change in relative 16S rRNA transcript abundance of individual OTUs when comparing activated sludge samples from different seasons. OTUs having higher 16S rRNA gene abundance in a seasonal comparison are more intensely colored toward the respective season color. The size of each OTU is proportional to the average 16S rRNA gene abundance between the two compared seasons. OTUs outside the dashed box in the center of each plot change more than 1 log in relative abundance between the two seasons, while <1-log changes are within the dashed box.

FIG 3.

Each plot compares two seasons ratio of 16S rRNA gene composition as well as the 16S rRNA transcription composition to discern how much variation occurs between successive seasons (a to d) and 6-month intervals (e and f). Each circle represents an OTU plotted with the average DNA percent abundance representing their size. Both axes represent a logarithmic ratio for the OTU percent abundance based on 16S rRNA gene (x axis) and 16S rRNA transcript (y axis). The intensity of the color scales proportionally with the OTUs’ higher abundance season. If an OTU is always within <1-log change based on the16S rRNA gene, it is at a constant abundance. If an OTU is always within <1-log change based on the 16S rRNA transcript, it is at a constant activity. There are 119 OTUs which meet both criteria for every plot and are within the dashed square.

Figure 3a to d show successive seasonal comparisons, while Fig. 3e and f show 6-month intervals to emphasize seasonal variations. Successional seasons (Fig. 3a to d) on average share 92.5% ± 2.8% of the same 16S rRNA gene composition and 97.9% ± 0.6% 16S rRNA transcript composition. Therefore, 7.5% ± 2.8% of the 16S rRNA gene communities are unique to each season. Only 2.1% ± 0.6% of the 16S rRNA transcript compositions were unique in successive seasons. On average, 79.3% ± 5.1% of the 16S rRNA gene composition and 82.7% ± 7.3% of the 16S rRNA transcript composition did not change more than 1 log between successive seasons.

Regarding compositions in fall to spring (Fig. 3e), the communities share 87.4% 16S rRNA gene compositions and differ by 12.6% each. However, only 63.2% and 68.6% of the 16S rRNA gene composition remains at similar abundances (<1 log), while 24.2% and 18.9% significantly change between fall and spring. The 16S rRNA transcript compositions are 94.9% and 96.7% similar in fall and spring, with 5.1% and 3.3% unique compositional changes. Despite the similar taxonomic compositions, 41.4% of fall and 19.5% of spring 16S rRNA transcript compositions significantly change between seasons, leaving 53.5% and 77.1% of the compositions at similar percent abundances.

Summer and winter (Fig. 3f) appear more similar, with 95.5% and 95.0% overlapping 16S rRNA gene composition and 80.7% and 84.0% 16S rRNA gene compositions at similar percent abundances. Therefore, only 4.5% and 5.0% are unique to each season and 14.8% and 11.0% significantly change between summer to winter. There is less variation among 16S rRNA transcripts, with 97.9% and 98.3% shared transcript compositions and 85.2% and 90.6% shared composition at similar percent abundances. Summer and winter samples had only 2.1% and 1.7% unique 16S rRNA transcripts and 12.7% and 7.7% significant variation in relative transcript abundance.

We observed linearity of OTU 16S rRNA gene and transcript abundances when comparing fall and spring samples (P ≪ 0.05; R² = 45.8%). Summer and winter do not significantly correlate 16S rRNA genes to transcripts (P > 0.95). All four successional seasons (Fig. 3a to d) had statistically significant correlations between 16S rRNA genes and transcripts (P ≪ 0.05), with low R² values (averaging 20.7% ± 7.5%).

A total of 119 OTUs did not change more than 1 log in gene or transcript abundance throughout the year. These 119 OTUs comprise the following percent abundances based on 16S rRNA gene composition per season: 69.6% in summer, 59.0% in fall, 75.7% in winter, and 63.6% in spring. These 119 OTUs comprise the following percent abundances based on 16S rRNA transcript composition: 63.5% in summer, 49.1% in fall, 70.4% in winter, and 73.3% in spring. The top five most abundant OTUs from these 119 OTUs are also listed in Fig. 4. Additional seasonally unique and significantly changing OTUs per season are listed in Tables S2 and S3.

FIG 4.

Triple Venn diagram representing the intersections of the three criteria for the consistent core microbiome. The three criteria were constant abundance (in blue; <1-log change in relative abundance throughout the year based on 16S rRNA genes shown in Fig. 3), constant activity (in yellow; <1-log change in relative abundance throughout the year based on 16S rRNA transcript shown in Fig. 3), and growth in activated sludge (in red; >1-log change in absolute abundance compared to influent wastewater as shown in Fig. 5). The top three OTUs based on relative abundance averaged across seasons and during the react cycles (average of R1 and R2) are listed for OTUs which meet one or two criteria, while the top five OTUs which meet all three criteria are listed. There are a total of 179 constant-abundance OTUs, 172 constant-activity OTUs, and 275 proliferating OTUs. A total of 108 OTUs met all three criteria as the consistent core microbiome.

Comparison of influent wastewater community and activated sludge.

To differentiate OTUs growing in activated sludge from OTUs passing through with influent wastewater, the absolute abundance of each OTU in the SF community was compared to the average absolute abundance during the react cycles (R1 and R2) using a mass balance of each OTU (equation S1). Figure 5 shows a comparison of absolute abundances of OTUs in the SF community to the combined react cycles (R1 and R2). OTUs increasing greater than 1 log from the SF community are OTUs which grow and recycle in activated sludge. OTUs with less than a 1-log difference in abundance do not significantly grow and will wash out. In Fig. 5, OTUs are represented by dots, and those OTUs within the shaded region do not significantly grow.

FIG 5.

A mass balance was performed on each OTU to discern which organisms are significantly growing and proliferating in activated sludge or passing through the system. The 16S rRNA gene relative abundance from sequencing is multiplied by the quantitative total abundance of 16S rRNA gene copies via qPCR to yield an absolute abundance of each OTU. Each dot represents a unique OTU which appeared in both SF and during the react cycle. The x axis is the abundance during SF, while the y axis is the average abundance throughout the year during R1 and R2. Nitrifying organisms are indicated. The red line shows the cutoff for OTUs which are greater than 1-log difference, the purple line indicates exactly 1-to-1 ratio, and the blue line shows a <1-log decrease cutoff. Anything falling between these bounds (purple shaded region) is considered at similar abundances entering the system and leaving the system at a nearly 1-to-1 ratio. Above this shaded region are OTUs which grew in abundance and are propagated in activated sludge.

A total of 506 OTUs were assessed in Fig. 5, but only 406 overlap between the reactor cycles. Of the 406 overlapping OTUs, 165 are in the shaded region which comprises 73.1% of the SF community. Although influent wastewater makes up, on average, a third of the reactor volume, these 165 OTUs constitute 11.2% of activated sludge. The region above the shaded area represents OTUs in the reactor which are growing in activated sludge. These 275 OTUs grow to represent 88.8% of the activated sludge community but are 25.7% of the influent wastewater community.

One hundred OTUs do not overlap between SF and the react cycles due to low sequencing reads. Sixty-six of these OTUs were in the SF community, comprising 1.2%. The remaining 34 OTUs were unique to activated sludge, representing 4.2% of the total activated sludge community.

The consistent core microbiome.

We found 108 OTUs meeting our three criteria for core microbiome. Figure 4 highlights 179 OTUs which exhibited consistent abundance (<1 log annually) and 172 OTUs with consistent protein synthesis activity (<1 log annually). A total of 119 OTUs exhibit both consistent 16S rRNA gene and transcript abundances. Of the 275 OTUs which were growing in activated sludge (Fig. 5), 108 had constant abundance and activity, meeting all three criteria for our core microbiome: growth, abundance, and activity. According to our 16S rRNA gene analysis, these OTUs comprise 58.7% of the activated sludge composition in summer, 51.5% in fall, 70.8% in winter, and 59.6% in spring. The same OTUs represent 60.2%, 46.9%, 66.5%, and 70.5% of the summer, fall, winter, and spring 16S rRNA transcript abundances, respectively. The top five OTUs of the consistent core microbiome are listed in Fig. 4 with their average annual relative abundances. Additionally, the top three OTUs fulfilling one or two criteria are listed in Fig. 4 with their respective average annual relative abundances.

Interestingly, 29.2% to 48.5% of the activated sludge microbiome composition based on 16S rRNA genes varies significantly in abundance or growth throughout the year. Accordingly, 29.5% to 53.1% of the 16S rRNA transcript composition of the activated sludge microbiome varies significantly in abundance.

Figure 6 highlights the top 50 most abundant OTUs according to our 16S rRNA gene amplicon sequencing results for samples collected during the react cycles. For the majority of OTUs, relative 16S rRNA gene abundances were proportional to the 16S rRNA transcript abundances with less than 1-log fold differences. While proportional, 16S rRNA transcripts copy numbers were generally higher than 16S rRNA gene copy numbers. The 50 OTUs listed in Fig. 6 comprise 56.7% ± 6.8% of the 16S rRNA gene community, yet they make up 67.5% ± 3.5% of the community based on 16S rRNA transcript sequencing. However, of the 50 most abundant OTUs, 10 had at least one season with more than a 1-log difference between gene and transcript copy numbers, while 5 showed higher than 1-log differences between gene and transcript copy numbers over multiple seasons. Leptotrichia was notably far less active in terms of 16S rRNA transcript abundance than the 16S rRNA gene abundance would have suggested (2.1% relative 16S rRNA gene abundance, compared to 0.1% relative 16S rRNA transcript abundance composition). Among the top 50 OTUs were several OTUs which declined in 16S rRNA gene abundance during winter or spring but maintained stable 16S rRNA transcript abundances and therefore had a high transcript-to-gene ratio. These OTUs include several uncharacterized organisms from Moraxellaceae, Myxococcales, “Candidatus Accumulibacter,” Chitinophagales, and Phaselicystis. While “Candidatus Accumulibacter” is a known phosphate-accumulating organism, the rest of these have yet-unknown roles in wastewater treatment facilities (14, 26).

FIG 6.

Relative abundances of the top 50 OTUs during the react cycles (average of R1 and R2) for all four seasons. The size of each circle is proportional to the relative sequence abundance. Panel a shows relative 16S rRNA gene sequence abundance, and panel b shows relative 16S rRNA transcript abundances. Panel c shows the ratio of transcripts to genes. Higher transcript relative abundances are indicated by solid circles, while higher gene abundances are indicated by open circles.

DISCUSSION

In this study, we investigated the response of the activated sludge microbiome in three full-scale SBRs of a WWTP previously described to experience seasonal variations in composition and seasonal nitrification failure. We sampled inflow wastewater and activated sludge and applied high-throughput 16S rRNA transcript sequencing alongside RT-qPCR to break down the complex composition of the activated sludge microbiome into taxonomic groups of microorganisms that were not only abundant residents of a reactor core community but also metabolically active and growing during a full reactor cycle.

Activated sludge systems are designed to recycle settled solids so microorganisms have longer generation times. 16S rRNA gene amplicon sequencing based on activated sludge DNA extracts picks up inactive and dormant species and extracellular DNA. 16S rRNA transcript sequencing in conjunction with 16S rRNA gene sequencing has the potential to reveal OTUs which maintain a detectable amount of rRNA, indicating their metabolic capacity for instant de novo protein biosynthesis. Considering only our 16S rRNA gene sequence data, we would have identified 179 OTUs as activated sludge core microbiome members based on the relative abundance. By also sequencing 16S rRNA transcripts, we were able to extend the core microbiome definition to include 119 OTUs not only consistently abundant but also active in the sense that they cell contained detectable amounts of rRNA as an indicator for their preparedness to synthesize new proteins.

Quantitative PCR revealed that total 16S rRNA transcript copy numbers were about 2 to 3 orders of magnitude higher than 16S rRNA gene copy numbers (Fig. 1). This was expected, as most cells maintain thousands to tens of thousands of ribosomes when they are metabolically active. In the fall, 16S rRNA transcript copy numbers were significantly higher during all reactor cycles than in winter, spring, and summer. While 16S rRNA transcript copy numbers in spring and summer were not significantly different from each other, they were greater than 16S rRNA transcript copy numbers in the winter during the react cycles. As a consequence of aeration with cold air during the winter, we observed a decrease in 16S rRNA transcript copy numbers by up to 2 orders of magnitude. This indicates that the activated sludge microbiome reacts with a decrease in protein synthesis potential due to the exposure to cold air during aeration and the consequential drop in water temperature. We observed this phenomenon when quantifying 16S rRNA transcripts, not 16S rRNA gene copy numbers, which serve as a proxy for total cell counts. The observed decrease in protein synthesis potential in the activated sludge community did not affect total cell biomass.

The activated sludge composition did not significantly change throughout reactor cycles, yet it was significantly different from the microbial community in the influent wastewater (SF) (Fig. 2a). By calculating absolute abundances, we found that 25.7% of the influent wastewater OTUs replicate and grow to comprise 88.8% of activated sludge, yet 73.1% of influent passes washes out as 11.2% of the activated sludge composition.

Seasonal shifts in the activated sludge microbiome.

Based on Bray-Curtis dissimilarity, the activated sludge composition was found to vary most when 16S rRNA gene composition was compared with 16S rRNA transcript composition (32.2% variance explained), emphasizing the value of considering sequence 16S rRNA transcript in addition to 16S rRNA genes (18). The seasons during which activated sludge samples were collected were the second most differentiating principal component describing activated sludge composition, with a total of 18.2% of explained total variance. Fall and spring samples had the most distinct compositions (Fig. 2b). Summer and winter compositions were statistically more similar based on Bray-Curtis dissimilarity for 16S rRNA transcripts (Fig. 2b).

The sampling dates in fall and spring succeeded the longest periods of continually warm and continually cold weather conditions. While fall and spring compositions appeared to be distinct (Fig. 2b), 16S rRNA transcript copy numbers during fall and spring were relatively stable and slightly higher than the reactor’s samples in summer and winter. The transient weather conditions leading into summer and winter resulted in more similar compositions between these two seasons, but the lower 16S rRNA transcript copy numbers during mixed fill and react cycles demonstrated a lower protein synthesis potential as the activated sludge microbiome tries to acclimate to changing temperatures. Temperature is continually demonstrated to be a significant driving factor affecting activated sludge community compositions (4, 7); however, by including quantification of 16S rRNA transcript copy numbers, we demonstrated how microbial community activity is also impacted by seasonal temperatures. The decrease in transcript copy numbers at the onset of aeration in winter demonstrates this phenomenon because the sampled wastewater treatment plant experienced a sudden cold front the day before sampling. Air temperatures decreased from −8°C to −20°C during the day and from −18°C to −33°C during the night. Cold shock and recovery have been observed in lab scale, but we show this phenomenon affecting a full-scale WWTP (27, 28). As the global climate changes the likelihood of more drastic weather-impacted shifts in temperature will rise, emphasizing why it is important to better understand the boundaries of resilience of activated sludge microbial community activities to ensure stable process performance in the future (29).

Based on Bray-Curtis dissimilarity, activated sludge composition varied prominently with seasons, yet shifts between reactor cycles were less pronounced. As expected, the DW had the most dissimilar composition, but not significantly. 16S rRNA transcript composition was expected to show more variation across reactor cycles since rRNA has a shorter half-life than genomic DNA (30). Because generation times of most microorganisms are longer than the average duration of a complete cycle (SRT≫HRT [25]), composition did not change significantly once the influent wastewater mixed with activated sludge.

Activated sludge microbial community resilience and plant performance stability are continuously impacted by the influent wastewater. In this study, we observed the largest community dissimilarity (38.2% [Fig. 2a]) between the influent wastewater sampled during SF and the activated sludge microbial community. While the activated sludge community composition showed slight seasonal variations, we observed little variation throughout the whole year in the influent wastewater chemical and composition. This has also been observed by Newton et al. previously (5). Since Ibarbalz et al. (54) showed that influent wastewater impacts activated sludge microbial community structure, we used a mass balance approach to determine which OTUs were growing (or not) in our SBRs. Individual OTU growth rates were calculated similarly to the method of Saunders et al. (10) but using 16S rRNA gene copy numbers for total biomass.

Among the OTUs which were not growing during the react cycles of the reactors, we identified primarily human health-related bacterial taxa. These included taxa such as Leptotrichia (2.1% ± 1.0%), which is found as part of the oral and intestinal human microflora (31); Arcobacter (1.0% ± 0.9%), a taxon comprising enteropathogenic bacteria (32); and Acinetobacter (0.9% ± 0.5%), containing species related to human infections (33). These OTUs have also been observed in influent and effluent wastewater by Numberger et al. (34), based on relative sequence abundance. Using a mass balance approach to approximate absolute OTU abundances, we found that human health-related OTUs, while present, were not growing in the activated sludge reactors. One exception was Mycobacterium (0.9% ± 0.0%), known to comprise species causing tuberculosis and leprosy. Mycobacterium OTUs were present and growing in the SBR; however, Mycobacterium was less active during the winter, because we observed a lower protein synthesis potential (35).

The parallel sequencing of DNA and RNA in conjunction with 16S rRNA gene and transcript copy number quantification enabled us to categorize the identified OTUs in the activated sludge samples in several different categories based on their abundance, activity, and growth in the system. OTUs which were identified as growing but fluctuated in abundance and activity by more than 1 log comprised members of the taxon Kouleothrix (1.7% ± 1.6%), uncultured OTUs belonging to the phylum Chloroflexi (1.0% ± 1.3%), and Gordonia (0.6% ± 0.5%). Kouleothrix has been described to be associated with sludge bulking (36), while members of the diverse phylum Chloroflexi include microorganisms capable of degrading halogenated hydrocarbons as well as filamentous microorganisms that are part of activated sludge flocs (37). Members of the taxon Gordonia have been described to be capable of degrading various xenobiotic compounds (38). Other floc-forming organisms of the genus Zoogloea fluctuated in abundance throughout the seasons, but within a lower range. Zoogloea is regarded as a ubiquitous taxon among wastewater treatment plants, with relative sequence abundance of typically around 0.89% (4). However, we observed the highest relative 16S rRNA gene abundance of Zoogloea in the fall at 0.56% and the lowest in the spring at 0.1%. Interestingly, based on relative 16S rRNA transcript abundances, Zoogloea comprised 1.77% in the fall and 0.55% in the winter. These observed changes in abundance and activity of floc-forming microbial taxa in activated sludge may contribute to seasonal variations in plant process performances by impacting activated sludge aggregate structure and stability.

Forty-eight OTUs identified as growing in the batch reactors, which made up an average of 4.95% of the bulk activated sludge community, also had a high protein synthesis potential but fluctuated in abundance by more than 1 log between seasons. The most abundant among these identified OTUs was Haliangium (0.87% ± 1.2%). Haliangium is detected in Carrousel oxidation ditch systems and might be involved in denitrification (39, 40). We do not know much about denitrifiers in SBRs, but since the WWTP undergoes seasonal nitrification failure, the abundance of denitrifiers might be affected. It is noteworthy that among the top 50 OTUs in Fig. 6, only Nitrotoga, a nitrite-oxidizing bacterium, is present, while Nitrosomonas, which is the predominant ammonia oxidizer, is far lower in abundance, at 0.20% ± 0.20%.

One objective of this study was to identify OTUs which belong to the activated sludge core microbiome in the sampled treatment plant. A total of 108 OTUs met our three criteria of abundance, activity, and growth. Most of these OTUs are typical members of activated sludge microbial communities, such as Tetrasphaera (2.9% ± 1.3%) (41), which is a known phosphate-accumulating microorganism, and Saprospiraceae (6.6% ± 3.5%), filamentous bacteria with a still-unknown role in activated sludge (39). Saprospiraceae was the highest-abundance OTU, with stable 16S rRNA gene and transcript abundances throughout the year, while our previous study noted it as the second most abundant OTU (3). Interestingly, some of the most ubiquitous OTUs recently identified by a survey of the Global Water Microbiome Consortium were not present in our SBRs (4). For example, we did not find evidence for the presence of the recently isolated organism Casimicrobium huifangae, which might play a role in phosphate removal and nitrate reduction. Our activated sludge communities also did not contain any members of the genus Nitrospira (42). While the Global Water Microbiome Consortium reported on the ubiquitous presence of Nitrospira in a global survey of WWTPs (4), the most abundant known nitrite oxidizers in the SBR samples in this study were Nitrotoga (0.2% ± 0.2%) and Nitrobacter (0.3% ± 0.2%) (43). However, the global assessment of activated sludge bacterial communities mostly featured conventional continuously stirred tank reactor configurations. Nitrosomonas (0.2% ± 0.2%) was the most abundant known ammonia oxidizer in our system. All three known nitrifying taxa present in the sampled treatment plant in this study, Nitrosomonas, Nitrotoga, and Nitrobacter, met our three criteria for consistent core OTUs (Fig. 5), based on consistent abundance, growth, and protein synthesis potential, despite seasonal nitrification failure (3). All three known nitrifiers remaining active and continuing to grow despite the seasonal nitrification failure raises the question of whether these species are metabolically more flexible than known and capable of switching when wastewater temperatures decrease.

Activated sludge presents one of the most diverse and dynamic engineered microbial ecosystems. This report presents a detailed analysis of the influent wastewater and activated sludge microbiome during the stages of a reactor cycle. By quantifying and sequencing the 16S rRNA genes and transcripts, we identified bacterial taxonomic groups based on their changes in abundance, protein synthesis potential, and continued growth in SBRs. This study improves our understanding of the microbial ecology of activated sludge and emphasizes the need to move from describing activated sludge compositions by relative sequence abundances to an absolute quantitative assessment incorporating activity of individual core taxa, if we want to better understand their role in community functions and resilience to perturbation in the future.

MATERIALS AND METHODS

Wastewater treatment plant.

Brainerd Wastewater Treatment Facility has three full-scale SBRs averaging 6-h-long cycles. Typically, 1 h of static filling (SF), 1 h of mixed fill (MF), 2 h of aerobic react cycles, 1 h of settling, and one final hour for decanting, wasting excess activated sludge, and sitting idle. The final hour for decanting, wasting sludge, and sitting idle was not sampled. The program is set to fill the reactor up to a third with influent wastewater and two-thirds with activated sludge before beginning the react cycles. Additional details of the plant configuration and standard operations have been described by Johnston et al. (3).

Sample collection.

The sample trip occurred within 10 days at the start of each season (1 July 2017, 3 October 2017, 27 December 2017, and 27 March 2018). Samples were collected every 10 min from each SBR cycle. The SBRs are equipped with a faucet which draws from an individual reactor. These faucets were used to collect samples from the reactors. The faucets were drawn for at least 30 s to ensure a fresh sample not idle in the pipes.

Five minutes prior to settling, we collected 20 liters of activated sludge in a carboy to simulate the settling occurring in the reactor. From the carboy, samples were collected from the settled sludge at the bottom using the carboy’s spigot valve, while grab samples were collected from the top half to represent decanted wastewater.

Sample aliquots in 2-ml microcentrifuge tubes were rapidly frozen to approximately –72°C in a bath of dry ice and ethanol. No RNA stabilization solutions were used since preliminary testing determined them to be less consistent than rapid freezing (data not shown). Samples were stored at –80°C and thawed prior to nucleic acid extractions or chemical analysis.

The transcriptional responses for 16S rRNA were quantified at every time interval. The 16S rRNA gene quantities were determined only at selected times. These time points represent the midpoint for static fill (30 min), mixed fill (80 min), the beginning of react cycle R1 (140 min), the ending of react cycle R2 (220 min), and the midpoint time for the decanted and settled sludge during settling (270 min). These six time points, for the triplicate reactors across the seasons, were used for 16S rRNA amplicon sequencing analysis to compare the 16S rRNA transcript and 16S rRNA gene abundances. These time points are abbreviated SF (static fill), MF (mixed fill), R1 (react 140 min), R2 (react 220 min), SS (settled sludge), and DW (decanted water).

Nucleic acid extraction and reverse transcription.

Frozen samples were centrifuged for 5 min at 13,000 × g to thaw and then resuspended and transferred into lysis buffer for either DNA or RNA extraction. Cells were disrupted using the FastPrep-24 5G homogenizer (Santa Ana, CA) for 40 s at 6 m/s. The DNA extractions were performed using 50 μl of homogenized sample and followed the protocol for the MP Biomedicals (Santa Ana, CA) FastDNA Spin kit for soil.

RNA extractions were performed using 250 μl of sample following the protocol for the Quick-RNA fecal/soil microbe microprep kit by Zymo Research (Irvine, CA). During the final step to elute RNA from the filter, 50 μl of DNase/RNase-free water was used to create aliquots of the extracted RNA. After RNA extraction, DNA was removed using the TURBO DNA-free kit from Invitrogen (Carlsbad, CA). The incubation of the DNase enzyme was increased from 30 to 40 min before inactivation to ensure removal of residual DNA. Finally, reverse transcription was performed using the PrimeScript RT-PCR kit from TaKaRa (Kusatsu, Japan). The incubation step was increased from 15 to 20 min before inactivation to ensure that reverse transcription had completed. During protocol development, each step was confirmed using qPCR/RT-qPCR for two functional genes (16S rRNA and amoA) as well as DNA and RNA high-sensitivity fluorometric assays on a Qubit 4 fluorometer by Invitrogen (Carlsbad, CA) as shown in the supplemental material (Fig. S1 to S3).

qPCR and RT-qPCR.

Quantification was performed on a Bio-Rad CFX Connect real-time PCR instrument (Hercules, CA). Each reaction mixture contained 10 μl of PCR-grade water (Ambion Inc., Foster City, CA), 12.5 μl of 2× SsoFast EvaGreen Supermix (Bio-Rad Laboratories, Hercules, CA), 1.25 μl of 10-mg/liter bovine serum albumin solution (Millipore Sigma, St. Louis, MO), 0.5 μl of forward primer (0.5 μM), 0.25 μl of reverse primer (0.5 μM), and 0.5 μl of template (c)DNA. The primer set used targeted the 16S rRNA genes/transcript (V3 region 338-F [CCTACGGGAGGCAGCAG] and 518-R [ATTACCGCGGCTGCTGG] [44]), which served as a proxy for total biomass and protein synthesis capability. For statistical analysis, when a value fell below the detection limit after repetitive attempts, a value of half the detection limit was used (45). The average amplification standard had an efficiency of 92.9% ± 6.9% and R² coefficient of 0.968 ± 0.041.

Sequencing and data analysis.

Sequencing was performed by the University of Minnesota’s Genomic Core using V1-V3 primers for 16S rRNA based under conditions described by Johnston et al. (3, 46). Sequencing analysis was performed using DADA2 in RStudio (47). Default settings were used unless noted with truncLen=c(290,270) to remove low-quality ends. Reads under 200 bp were removed [minLen=c(200,200)]. Additional settings were maxMismatch = 1 and minOverlap = 20, which merges only ≥20 overlapping base pairs with 1 mismatching base pair. The distribution of sample reads is shown in Fig. S4.

Operational taxonomic units (OTUs) referenced SILVA rRNA SSU 132 (48) at the 97% cutoff. DADA2 classified 1,002 OTUs at the genus level. The average sample finished with 44,601 ± 18,580 merged reads. Seven samples were poorly amplified, with a total of less than 1,000 reads, and were removed from the data set (Table S1).

Statistical analysis.

Statistical analysis and graphing were performed using RStudio version 3.6.0. Bray-Curtis dissimilarity was performed using the vegan package to calculate the distances and percent variance explained (49, 50). Adonis tests were used to determine the significance of ordination clusters. Analysis of variance (ANOVA) and regression analysis were used to determine the relationship between transcript abundances over time. Student t tests were used to determine significant differences across gene/transcript quantifications.

Consistent core microbiome requirements.

The three criteria we developed to determine the most stable organisms were abundance, activity, and growth. (i) Abundant OTUs were detectable each season (using an average of R1 and R2 for the three SBRs) and did not vary in relative 16S rRNA gene sequence abundance more than 1 log across all seasons. (ii) Active OTUs were detectable each season (using an average of R1 and R2 for the three SBRs) and did not vary in 16S rRNA transcript relative abundance more than 1 log across all seasons. The 1-log cutoffs were selected because microbial growth and diversity are log-normally distributed and meaningful biological differences often only occur over logarithmic increments (51). Additionally, we used these 1-log cutoffs since our previous works that focused on dynamic OTU abundance shifts over time have shown that OTU abundances significantly correlated with seasonal water temperature variations (annual sine wave pattern), regardless of the amplitude of the actual temperature change. In this study, we wanted to focus on larger-scale seasonal abundance variations of OTUs by selecting a 1-log cutoff to be applied to log-transformed data (52, 53). (iii) Growing OTUs measured a >1-log increase in absolute abundance when comparing the influent wastewater to the activated sludge community (using an average of R1 and R2 for the three SBRs). The absolute abundance was calculated by multiplying the relative 16S rRNA gene sequence abundance by the total 16S rRNA gene copy number obtained by qPCR for each of the sampled sequencing batch reactor cycle phases. A more detailed equation and explanation can be found in the supplemental material (equation S1).

Data availability.

Sequences are available from the NCBI Sequence Read Archive with accession number PRJNA591266.