Population Structure and Morphotype Analysis of “Candidatus Accumulibacter” Using Fluorescence In Situ Hybridization-Staining-Flow Cytometry

As one group of the most important functional phosphorus removal organisms, “Candidatus Accumulibacter,” affiliated with the Rhodocyclus group of the Betaproteobacteria, is a widely recognized and studied PAO in the field of biological wastewater treatment. The morphotypes and population structure of clade-level “Candidatus Accumulibacter” were studied through novel FISH-staining-flow cytometry, which involved denitrifying phosphorus removal (DPR) achieving carbon and energy savings and simultaneous removal of N and P, thus inferring the different denitrifying phosphorus removal abilities of these clades. Additionally, based on this method, in situ quantification for specific polyphosphate-accumulating organisms (PAOs) enables a more efficient process and more accurate result. The establishment of FISH-staining-flow cytometry makes cell sorting of clade-level noncultivated organisms available.

ABSTRACT

“Candidatus Accumulibacter” is the dominant polyphosphate-accumulating organism (PAO) in denitrifying phosphorus removal (DPR) systems. In order to investigate the community structure and clade morphotypes of “Candidatus Accumulibacter” in DPR systems through flow cytometry (FCM), denitrifying phosphorus removal of almost 100% using nitrite and nitrate as the electron acceptor was achieved in sequencing batch reactors (SBRs). An optimal method of flow cytometry combined with fluorescence in situ hybridization and SYBR green I staining (FISH-staining-flow cytometry) was developed to quantify PAOs in DPR systems. By setting the width value of FCM, bacterial cells in a sludge sample were divided into three groups in different morphotypes, namely, coccus, coccobacillus, and bacillus. Average percentages that the three different PAO populations accounted for among total bacteria from SBR1 (SBR2) were 42% (45%), 14% (13%), and 4% (2%). FCM showed that the ratios of PAOs to total bacteria in the two reactors were 61% and 59%, and the quantitative PCR (qPCR) results indicated that IIC was the dominant “Candidatus Accumulibacter” clade in both denitrifying phosphorus removal systems, reaching 50% of the total “Candidatus Accumulibacter” bacteria. The subdominant clade in the reactor with nitrite as the electron acceptor was IID, accounting for 31% of the total “Candidatus Accumulibacter” bacteria. The FCM and qPCR results suggested that clades IIC and IID were both coccus, clade IIF was coccobacillus, and clade IA was bacillus. FISH analysis also indicated that PAOs were major cocci in the systems. An equivalence test of FCM-based quantification confirmed the accuracy of FISH-staining-flow cytometry, which can meet the quantitative requirements for PAOs in complex activated sludge samples.

IMPORTANCE As one group of the most important functional phosphorus removal organisms, “Candidatus Accumulibacter,” affiliated with the Rhodocyclus group of the Betaproteobacteria, is a widely recognized and studied PAO in the field of biological wastewater treatment. The morphotypes and population structure of clade-level “Candidatus Accumulibacter” were studied through novel FISH-staining-flow cytometry, which involved denitrifying phosphorus removal (DPR) achieving carbon and energy savings and simultaneous removal of N and P, thus inferring the different denitrifying phosphorus removal abilities of these clades. Additionally, based on this method, in situ quantification for specific polyphosphate-accumulating organisms (PAOs) enables a more efficient process and more accurate result. The establishment of FISH-staining-flow cytometry makes cell sorting of clade-level noncultivated organisms available.

INTRODUCTION

The excessive discharge of wastewater rich in nitrogen (N) and phosphorus (P) pollutants is the main cause of eutrophication (1). Efficient removal of nitrogen and phosphorus from wastewater has become an urgent problem. An enhanced biological phosphorus removal (EBPR) process has been widely used in wastewater treatment due to its economical and environmentally friendly characteristics. Polyphosphate-accumulating organisms (PAOs) play an important role in the EBPR process. Compared with traditional EBPR, the denitrifying phosphorus removal (DPR) process significantly reduces the demand of a carbon source and energy, which has caused worldwide concerns. Denitrifying polyphosphate-accumulating organisms (DPAOs) take up volatile fatty acids (VFAs) to store as poly-β-hydroxyalkanoates (PHAs) by decomposing polyphosphate under anaerobic conditions. Meanwhile, phosphorus is released into mixed sludge liquid. Under anoxic conditions, the energy for P uptake is provided by PHA oxidation with nitrate or nitrite as the electron acceptor, achieving simultaneous removal of N and P, saving energy costs, and reducing external carbon source addition (2–4). As the important functional microorganism, the abundance and bioactivity of DPAOs determine the efficiency of N and P removal performance (5).

To date, PAOs cultivated in isolation were rarely reported. Molecular biological methods independent of pure culture become an effective way for quantitative and qualitative analysis of PAOs. Previous studies reported PAOs mainly affiliated with the Rhodocyclus group of the Betaproteobacteria, which were named “Candidatus Accumulibacter phosphatis” (6). From laboratory-scale to full-scale EBPR systems, “Candidatus Accumulibacter” was a numerically dominant PAO (7, 8). Real-time quantitative PCR (qPCR) and fluorescence in situ hybridization (FISH) are the primary approaches for quantitative analysis of PAOs. He et al. (9) investigated the abundances and distributions of the total “Candidatus Accumulibacter” bacteria and five “Candidatus Accumulibacter” clades from three laboratory-scale and nine full-scale EBPR samples and two full-scale non-EBPR samples through designing specific primers for “Candidatus Accumulibacter” 16S rRNA genes and polyphosphate kinase 1 (ppk1) genes. Based on qPCR, Camejo et al. (10) designed new specific primers targeting ppk1 genes in fourteen clades of the “Candidatus Accumulibacter” lineage to detect the clade-level population structure of “Candidatus Accumulibacter” in microaerobic reactors. In addition, the 16S rRNA FISH probes were designed to analyze two different types of “Candidatus Accumulibacter.” Results indicated that clade IA, bound to the Acc-I-444 probe, was a rod and could reduce nitrate to nitrogen gas. However, clade IIA, bound to the Acc-IIA-444 probe, was coccobacillus and did not have denitrifying abilities (11, 12).

However, since the copy numbers of 16S rRNA genes and certain functional genes in each genome are variable (13–15), multicopy genes bring many problems to the quantitative analysis of microorganisms using 16S rRNA primers, which makes diversity estimation and structural analysis difficult. Research indicated that the “Candidatus Accumulibacter” genome had two copies of the rrn operon (16) and bacterial genomes had 4.1 copies (10, 17, 18) or 3.6 copies (19, 20) of the rrn operon. Thus, multicopy genes should be considered in qPCR-based analysis of DPAOs.

In addition, FISH using specific probes to mark functional species is an easily available detection method (21, 22). Early studies of flow cytometry (FCM) mainly focused on bacterial structure and bioactivity (23–25) and, subsequently, on detecting population dynamics of bacteria in drinking water (26–28). However, few studies on quantitative detection of specific PAOs in wastewater treatment systems using FCM have been reported despite its low detection error (standard deviation [SD] < 5%) and efficient process (minutes) (29). Considering the limit of quantifying bacterial clusters in different morphotypes by using FISH and of specifically analyzing strains from environmental samples through FCM alone, a combination of FCM and FISH can provide a new alternative approach in quantifying and even in specific cell sorting. Based on this novel method and qPCR, analyzing “Candidatus Accumulibacter” clade community structure in sequencing batch reactor (SBR) systems would be more conducive to illustrating the microorganisms’ mechanism in operating wastewater treatment plants (WWTPs).

This study focused on the quantitative analysis of PAOs in denitrifying phosphorus removal systems using nitrate or nitrite as electron acceptors, which aimed to (i) determine the optimal quantitative method by comparing the accuracy of FISH-flow cytometry (fluorescence in situ hybridization combined with FCM) and FISH-staining-flow cytometry (fluorescence in situ hybridization and dye staining combined with FCM), (ii) determine the relative abundance and morphological characteristics of “Candidatus Accumulibacter” clades, and (iii) analyze the performance of denitrifying phosphorus removal systems based on the quantitative results of FCM and qPCR.

RESULTS AND DISCUSSION

Long-term performance of two SBRs.

To avoid the inhibition of overaccumulated nitrate and nitrite, the whole operational period was divided into three phases and lasted for 112 days (Table 1). During the first phase (1 to 27 days), NO₃⁻-N was just added once per cycle, and its average mixed concentration was 23 mg/liter in SBR1. At the end of the anoxic stage, occasional P was not completely removed, which was possibly attributed to the insufficient nitrates as electron acceptors for P uptake and carbon consumption rate (Fig. 1a). Oxygen as the electron acceptor in the aerobic condition ensured further P removal and avoided the harm of long-term operation under the anaerobic-anoxic condition to DPR systems (30, 31). In phase II (28 to 60 days), nitrate in SBR1 was increased to an average of 34 mg/liter and added only once per cycle. P removal efficiencies stabilized above 90%, while NO₃⁻-N in effluent was relatively high (Fig. 1). In order to avoid the inhibition of nitrite in SBR2, nitrite was added twice per cycle with a single dose average of 19 mg/liter. P released in the anaerobic period gradually increased and could be removed at the end of the anoxic stage with an anoxic P removal efficiency of more than 90% (Fig. 1c). However, nitrite cannot be completely denitrified, resulting in a high concentration of nitrite in the effluent (Fig. 1d). Considering that long-term residual nitrate may damage the DPR system (32), nitrate was added twice per cycle in phase III with an average of 22 mg/liter for each dosing. Nevertheless, during days 75 to 90, the anaerobic P release amount significantly decreased in both SBRs (Fig. 1a and c). Insufficient organic matter conflicted with the growth of PAOs needing more carbon sources, which resulted in the decrease of PHA synthesis and P release under anaerobic conditions (33). With the increase of 200 to 300 mg/liter chemical oxygen demand (COD) for each cycle after day 90, anaerobic P release obviously increased.

TABLE 1.

Experimental scheme for enrichment of denitrifying polyphosphate-accumulating organisms^d

Phase	SBR operation				COD (mg/liter)	PO43−-P (mg/liter)	C/P	No. of pulsesa		Concentration (mg NOX-N/liter)b
	Anaerobic (min)	Anoxic (min)	Aerobic (min)	Cycle (h)				NO3−	NO2−	NO3−	NO2−
I (1–27 days)	120	240	60	8	200	10	20	1	1	23	19
II (28–60 days)	120	240	60	8	200	10	20	1	2	34	19
III (61–90 days)c	120	240	60	8	200	10	20	2	3	22	14
III (91–112 days)c	60	300	60	8	300	10	30

^aNumber of times a NO_X-N pulse was added.

^bSingle pulse concentration of NO_X-N.

^cPhase III was divided into two parts.

^dThe DO concentration for the entire operation was 2 ± 0.5 mg NO_X-N/liter, and the pH was 7.5 ± 0.5.

FIG 1.

N and P removal over the 112-day operational period. (a) P removal of SBR1. (b) NO₃⁻-N removal of SBR1. (c) P removal of SBR2. (d) NO₂⁻-N removal of SBR2. d, days.

After influent COD adjusted from 200 mg/liter to 300 mg/liter, anaerobic-synthetized PHAs rose from being below 15 mg/liter to an average of 45 mg/liter (see Fig. S1 in the supplemental material). That poly-β-hydroxyvalerate (PHV) accounted for more than 80% of total PHAs could result from sodium propionate as the carbon source in the DPR system (34, 35). This indicated that carbon insufficiency seriously affected PHA synthesis and further hindered anaerobic P release, thus influencing the enrichment of PAOs. The typical operation cycle of SBRs is shown in Fig. S2 in the supplemental material.

Population structure of “Candidatus Accumulibacter.”

Genomic DNA was extracted from the freeze-dried sludge samples of SBR1 and SBR2 operated at the 110th day. Total bacteria and total “Candidatus Accumulibacter” bacteria were quantified based on 16S rRNA genes, and “Candidatus Accumulibacter” clades were quantified based on the ppk1 gene, where the bacterial 16S rRNA totals in the SBR1 and SBR2 samples were 2.52 × 10¹² copies/g dried sludge and 6.26 × 10¹¹ copies/g dried sludge, respectively. As for “Candidatus Accumulibacter,” the 16S rRNA gene abundance in the two reactors were 7.27 × 10¹¹ copies/g dried sludge and 1.48 × 10¹¹ copies/g dried sludge (Fig. 2a). Due to the controversy regarding the average copy numbers of 16S rRNA genes contained in a single bacterial genome, 3.6 copies (19, 20) and 4.1 copies (9, 17) were both used for the conversion of bacterial cell abundance in this study. The genome of “Candidatus Accumulibacter” contained 2 copies of the 16S rRNA gene and 1 copy of ppk1; thus, the abundance of ppk1 can represent the cell abundance of “Candidatus Accumulibacter” (10, 16, 18). Studies for “Candidatus Accumulibacter” have shown that the “Candidatus Accumulibacter” lineage is comprised of two types (type I and type II), each consisting of numbers of monophyletic clades (9, 10). However, among these clades, IIH and III were seldom discovered and discussed in full-scale WWTPs (18). He and McMahon, systematically reporting on “Candidatus Accumulibacter,” indicated that type I, excluding IA, was mostly from natural habitats (36). Additionally, clades IB through IE, clade IIE, and clade IIG were rare and almost had not been monitored (37–39). Mostly, clade IA and clades IIA through IIF (except IIE) were reported as the most common dominant “Candidatus Accumulibacter” clades in various kinds of WWTPs, especially in EBPR and the anaerobic-anoxic-oxic (A2O) processes.

FIG 2.

(a) Quantification of total bacteria, “Candidatus Accumulibacter” bacteria, and its clades by qPCR. ppk1 gene, 1 copy/cell in “Candidatus Accumulibacter” clades; 16S rRNA gene, 2 copies/cell in “Candidatus Accumulibacter,” 3.6 copies/cell and 4.1 copies/cell in all bacteria are considered in this study. (b) Redundancy analysis (RDA) diagram. The relationships between “Candidatus Accumulibacter” (Acc) clades and water physicochemical properties under different operational phases. PT, pulse times (number of times a pulse was added); INF. NO3, influent NO₃⁻-N; ANO. P, anaerobic phosphorus release; EFF. P, effluent phosphorus; INF. NO2, influent NO₂⁻-N; TEMP, temperature. The correlation between each species and environmental variable can be predicted by the length of the arrows projected onto the imaginary line running in the direction of a specific arrow.

In this study, the cell abundances of “Candidatus Accumulibacter” clades are shown in Table 2. Dominant “Candidatus Accumulibacter” clade IIC in SBR1 accounted for 52.14% of total “Candidatus Accumulibacter” bacteria and reached 1.9 × 10¹¹ cells/g dried sludge. The abundance of IIC (ppk1 excluding operational taxonomic unit [OTU] NS D3) as the secondary dominant clade was 1.39 × 10¹⁰ cells/g dried sludge and accounted for 3.82% of total “Candidatus Accumulibacter” bacteria. In contrast, the proportions of clades IA, IIA, IIB, IID, and IIF were lower. In SBR2, the abundance of IIC was 3.17 × 10¹⁰ cells/g dried sludge and accounted for 42.94% of total “Candidatus Accumulibacter” bacteria followed by IID (30.89%) and IIF (14.95%). Previous studies showed that clades IIC and IID performed well in denitrifying phosphorus removal in a continuous flow process treating domestic sewage (40). Moreover, clades IIC and IIF enriched in an SBR with nitrate as an electron acceptor can also achieve good denitrification and phosphorus removal performances (41).

TABLE 2.

Cell abundance and ratio of “Candidatus Accumulibacter” clades by qPCR in this study

Reactor	Parameter	Clade of “Candidatus Accumulibacter”							Total “Candidatus Accumulibacter” bacteriaa/total bacteriab (%)
		IA	IIA	IIB	IIC	IIC (ppk1 excluding OTU NS D3)	IID	IIF
SBR1	Cells/g dried sludge	7.47 × 109	4.03 × 108	2.10 × 106	1.90 × 1011	1.39 × 1010	1.06 × 1010	5.08 × 109	51.93∼59.2
	Clades/total “Candidatus Accumulibacter” bacteriaa (%)	2.06	0.11	0.00058	52.14	3.82	2.91	1.40
SBR2	Cells/g dried sludge	2.46 × 109	8.56 × 108	1.87 × 106	3.17 × 1010	4.14 × 109	2.28 × 1010	1.10 × 1010	42.5∼48.4
	Clades/total “Candidatus Accumulibacter” bacteriaa (%)	3.33	1.16	0.00252	42.94	5.60	30.89	14.95

^aTotal “Candidatus Accumulibacter” cell concentrations of SBR1 and SBR2 were 3.64 × 10¹¹ cells/g dried sludge and 7.39 × 10¹⁰ cells/g dried sludge.

^bTotal bacterial cell concentrations of SBR1 and SBR2 were shifted to 6.14 × 10¹¹∼7 × 10¹¹ cells/g dried sludge (SBR1) and 1.53 × 10¹¹∼1.74 × 10¹¹ cells/g dried sludge (SBR2).

It is worthwhile to notice that the dominant “Candidatus Accumulibacter” clade was IIC in both SBR1 and SBR2. Previous studies have shown that clade IIC has the ability to reduce nitrate, but denitrifying contributions cannot be only attributed to this clade (10). Genomic drafts showed that IIC lacked the nor gene and the nos gene and did not have a complete denitrifying metabolic pathway (42). Thus, there may be other flanking denitrifying bacteria involved in the process of denitrifying and phosphorus removal in DPR systems (31).

Changes of physiochemical parameters were always followed by the variation of microbial community structure (43, 44). Redundancy analysis (RDA) showed that the first axis explained 37.11% of the total variation between species and water physicochemical properties and 26.42% for the second ordination axis (Fig. 2b). The pH, number of pulses, COD, and temperature have greatly affected the abundances of IA, IIC, IID, and IIF, among which COD and temperature were crucial factors (45, 46). They indicated that an appropriate carbon concentration and dosing mode essentially determined DPR performance considering its provision of the main substrate source. A higher temperature was advantageous to accelerate the metabolism and growth speed of these clades (47). Anaerobic-released phosphorus positively relating with IID and IIF showed that the increases of these two clades were always followed by the increase of anaerobic P release. However, as one of the most significant negative inhibiting factors, high dissolved oxygen (DO) was probably more beneficial to other “Candidatus Competibacter” bacteria competing for resource, thus restraining accumulation of clades IA, IIC, IID, and IIF (48). Additionally, P increasing in effluent always means deterioration in DPR. Influent NO₂⁻-N has significant negative correlation to all clades, which was exactly contrary to influent NO₃⁻-N (33). Unstable DPR induced by nitrite accumulation would make it more difficult to enrich “Candidatus Accumulibacter” clades. In this study, with the nitrite decreased and the number of pulses increased along the entire operation, all clade abundances increased, which was consistent with Fig. 2b. Additionally, appropriate high nitrate promoted the enrichment of “Candidatus Accumulibacter” clades.

Quantitative analysis of PAOs by FISH-staining-flow cytometry.

Distinguished from counting by using SYBR or propidium iodide (PI) dye (49, 50), in this study, FISH-FCM and FISH-staining-FCM (fluorescence in situ hybridization conjugated with dye staining and FCM) were compared in detecting PAOs (see Fig. S3 in the supplemental material). Based on FISH-staining-flow cytometry, bacteria containing complete cell structure and double-stranded DNA could be distinguished from inanimate particles by staining with SYBR green I. Position distributions and proportions of PAOs and total bacteria in dot plot were available by backgating Cy5-positive cells in P3 (Fig. 3a and b). A total of 30,000 events were collected from the SBR1 sample, including cell fragments, inorganic particles, and bacterial cells with complete cell structure (Fig. 3a). Events in P1 (blue, yellow, and green dots) accounted for 67.7% of all events. Cells containing double-stranded DNA and stained with SYBR green I were gated as P1, which excluded non-double-stranded DNA events (red dots on both sides). Cells in P2 (yellow and green dots) were derived from P1 and accounted for 78.7% of P1. By gating P2, some events were removed from the detection limit, such as cell-free DNA and some fungus (blue dots on both sides). The ratio of Cy5-positive PAOs (green dots) in P3 was 61.3%, that is, PAOs accounting for 61.3% of the total bacteria in the SBR1 sample. Figure 3b showed that 30,000 events were collected from the SBR2 sample, and cells in P1 contained double-stranded DNA, which accounted for 75% of all collected events. Bacterial cells with consistent size in P1 were gated as P2, accounting for 79.5% of P1. Cy5-positive populations (P3) of the SBR2 sample were 58.9%. Details on the gating process are presented in Fig. S4 in the supplemental material.

FIG 3.

The position and proportion distribution of PAOs in dot plot by backgating. (a) SBR1 sample (61.3% positive Cy5). (b) SBR2 sample (58.9% positive Cy5).

Morphotype analysis of PAOs.

Based on FCM, bacterial cells of different sizes and types were identified by selecting different axis parameters during detection. Ratios of PAOs were 61.3% (SBR1) and 58.9% (SBR2) through FCM combined with SYBR green I staining and FISH. Parameter width means the time used by laser to excite a single cell. Different width values correspond to bacterial cells of different sizes. SX (X = 1, 2, 3) represents three different groups of total bacteria, and SX-1 (X = 1, 2, 3) represents three different groups of PAOs in Fig. 4. Proportions of PAOs accounting for total bacteria in the SBR1 sample were Group_x of PAOs/total bacteria (X = 1, 2, 3). Proportions of PAOs in total PAOs were Group_x of PAOs/total PAOs (X = 1, 2, 3), similar to the SBR2 sample (Table 3).

FIG 4.

The classification of bacterial populations during the quantitative analysis of SBR samples by flow cytometry. Events in P2 from the SBR1 and SBR2 samples were divided into three bacterial groups (S1 [B1], S2 [B2], and S3 [B3]) in a width-SSC-A plot. Proportions of positive Cy5 (S3-1 [B3-1], S2-1 [B2-1], and S1-1 [B1-1]) in each group were different, 80.3% in S3 (80.8% in B3), 69.7% in S2 (72.3% in B2), and 56.5% in S1 (57% in B1).

TABLE 3.

Quantitative result of groups of PAOs by FISH-staining-flow cytometry

Parameter	Groupx of SBR1 sample			Groupx of SBR2 sample
	Group 1	Group 2	Group 3	Group 1	Group 2	Group 3
Groupx of bacteria/total bacteria (%)a	74.70	20.20	4.78	79.60	17.50	2.62
Groupx of PAOs/groupx of bacteria (%)b	56.50	69.70	80.30	57	72.30	80.80
Groupx of PAOs/total bacteria (%)	42.21	14.08	3.84	45.37	12.65	2.12
Groupx of PAOs/total PAOs (%)	70.20	23.42	6.39	75.44	21.03	3.53

^aThe ratio of SX (BX) gated by FCM is Group_x of bacteria/total bacteria (%) from the SBR1(2) sample (X = 1, 2, 3).

^bThe ratio of SX-1 (BX-1) gated by FCM is Group_x of PAOs/Group_x of bacteria (%) from the SBR1(2) sample (X = 1, 2, 3).

Interestingly, Carvalho et al. (30) found that dominant bacteria enriched in a propionate-fed reactor with nitrate as the electron acceptor were bacilli, and the dominant bacteria in an acetate-fed reactor were cocci. Later, Oehmen et al. (12) found that rod morphotype bacteria, which bound to the Acc-I-444 probe (targeting “Candidatus Accumulibacter” type I), only existed in the propionate-fed reactor, and coccobacilli only bound to the Acc-II-444 probe (targeting “Candidatus Accumulibacter” type II) and were present in both propionate-fed and acetate-fed reactors. Based on previous studies, three different bacterial morphologies detected by FCM in this study possibly corresponded to cocci in S1/B1 (group 1), coccobacilli in S2/B2 (group 2), and bacilli in S3/B3 (group 3). qPCR indicated that the dominant clade in SBR1 with nitrate as the electron receptor was IIC, which accounted for 52.14% of total “Candidatus Accumulibacter” bacteria (Table 2). FCM-based quantification showed that PAOs of group 1 with the highest ratio were gated as S1 (Fig. 4). Therefore, clade IIC could be subordinated to group 1, which was coccus in small size. Dominant clades IIC and IID in SBR2 accounted for 42.94% and 30.89% of total “Candidatus Accumulibacter” bacteria, respectively, and were much greater than the PAOs of group 2 (21.03%). This suggested that IIC and IID were not in B2, but the coccus of group 1 in B1. FISH analysis also confirmed that the dominant bacteria in the two SBRs were cocci (see Fig. S5 in the supplemental material). Oehmen et al. (34) enriched denitrifying PAOs in a propionate-fed SBR with nitrite as the electron acceptor, and morphological characteristics of dominant bacteria were cocci, which was consistent with this study and indicated that IIC and IID were capable of reducing nitrite. IIF in SBR2 accounted for 14.95% of total “Candidatus Accumulibacter” bacteria, which was much higher than the proportion of PAOs in B3 (3.53%). This indicated that IIF was coccobacillus of group 2 in B2 and could reduce nitrite as well. Moreover, the proportion of IIC (ppk1 excluding OTU NS D3), accounting for 5.6% of total “Candidatus Accumulibacter” bacteria, was also higher than the proportion of PAOs in B3 (3.53%), indicating that IIC (ppk1 excluding OTU NS D3) was not bacillus of group 3 but coccus or coccobacillus.

It has been reported that clade IA, as the most important denitrifying bacteria in EBPR with nitrate as the electron acceptor, is a rod (12, 30, 41, 51). In this study, the ratios of clade IA in SBR1 and SBR2 were 2.06% and 3.33% (Table 2), respectively, slightly lower than the ratio of group 3. Therefore, clade IA was a rod morphotype PAO of group 3 in S3 (B3) detected by FCM, which could be denitrifying with nitrate.

Equivalence test and significance analysis.

To get a more accurate quantitation and verify the reliability of experimental data, an equivalence test was applied to the PAOs/total bacteria from one sludge sample quantified twice by FCM. Two independent quantifications in the SBR1 sample, 61.3% and 56% (Fig. 5), were substituted into equivalence test formulas (see the supplemental material), and the calculated confidence interval was [−0.062323, −0.044677] within [−0.1, 0.1]. For the SBR2 sample, two independent quantitative results were 58.9% and 50.3% (Fig. 5). The confidence interval [−0.095257, −0.077743] was within the tolerance interval [−0.1, 0.1]. It indicated that these data sets were equivalent, which truly reflected the relative abundance of PAOs. Accordingly, the more accurate average quantitative results obtained were 58.65% ± 3.75% (SBR1) and 54.6% ± 6.08% (SBR2).

FIG 5.

The second quantification of PAOs by flow cytometry (56% for positive sample 1 of SBR1; 50.3% for positive sample 2 of SBR2).

Comparison of PAOs quantitated by qPCR and FISH-staining-flow cytometry is shown in Table 4. Significance analyses for the quantifications acquired by the two methods showed that P = 0.579 (SBR1) and P = 0.237 (SBR2) with α = 0.05. The difference between quantification results determined by FCM and qPCR was not significant. An equivalence test for FCM data sets and significance analysis both confirmed the stability and sensitivity of FISH-staining-flow cytometry, which satisfied the quantitative requirements of PAOs in a complex activated sludge system.

TABLE 4.

Comparison of the quantitative results by two methods

Method	SBR1d		SBR2d
	PAO (%)	SD (%)	PAO (%)	SD (%)
FISH-staining-flow cytometrya	58.65	3.75	54.60	6.08
qPCR	59.2b	5.14	48.4b	5.90
	51.93c		42.5c

^aSYBR green I staining combined with probe PAO_MIX (Cy5).

^bThe chosen average 16S rRNA copy number of the bacterial genome was 4.1 copies.

^cThe chosen average 16S rRNA copy number of the bacterial genome was 3.6 copies.

^dSignificance values were as follows: P = 0.579 (SBR1), P = 0.237 (SBR2); α = 0.05.

To conclude, PAOs in DPR systems could be accurately quantified by novel FISH-staining-flow cytometry. Combined with the qPCR-based results, FCM could determine the different morphotypes of “Candidatus Accumulibacter” clades. These findings not only provide a new alternative efficient quantification method but also make it possible that cell sorting of clade-level noncultivated strains will greatly benefit the deep analysis of environmental functional microbes. PAOs sorted will be sequenced, especially denitrifying functional genes nirS, nirK, and narG in PAOs, to illustrate the relations between PAO and conventional denitrifying bacteria. Instead of conventional hybridization, which is known to be time-consuming work with optical instability probes, the dye and fluorescent antibodies as markers are always stable even when samples are sometimes exposed to sunlight. It is essential to find appropriate dye markers or protein antibodies as specific markers for targeted strains so that quantification and specific cell sorting by flow cytometry will be more available.

MATERIALS AND METHODS

Reactor setup and operation.

The seed sludge was derived from SBRs operated for 4 months with alternating anaerobic-aerobic conditions. Two sequencing batch reactors, SBR1 and SBR2, with an effective working volume of 4 liters were operated under anaerobic-anoxic-aerobic conditions at 25°C with nitrate and nitrite as electron receptors, respectively. The pH was controlled at 7.5 ± 0.5 by adding 0.2 mol/liter HCl or NaOH. The airflow meter controlled the aeration rate to keep DO concentration at 2 ± 0.5 mg/liter in the aerobic stage. Two liters of synthetic wastewater was fed per cycle with a hydraulic retention time (HRT) of 16 h. The concentration of mixed liquid suspended solids (MLSS) was maintained at 3,500 ± 500 mg/liter by discharging 167 ml of mixed liquid each cycle with a sludge retention time (SRT) of approximately 8 days. Two reactors were operated with 3 cycles each day, and each cycle of 8 h consisted of a 120-min (or 60 min) anaerobic period (including a 7-min fill period), a 240-min (or 300 min) anoxic period, a 60-min aerobic period, 40 min of settling, 3 min of decanting, and a 17-min idle period (Fig. 6 and Table 1).

FIG 6.

Cyclic operation scheme of SBRs.

By adjusting the dose of nitrate and nitrite, the whole operation period was divided into three phases to avoid the harm of nitrate or nitrite accumulation to the system (Table 1). During phase I, nitrate and nitrite were separately added only once at the beginning of the anoxic period. In phase II, nitrate was added once into SBR1 and nitrite was added twice into SBR2 with an interval of 1 h. In phase III, nitrate was added twice and nitrite was added three times per cycle with an interval of 1 h. Since a long-term operation with a low C/P ratio would inhibit the enrichment of PAOs, influent COD was increased to 300 mg/liter (C/P = 30) by adding sodium propionate, and anaerobic stirring duration decreased from 2 h to 1 h after the 90th day.

Synthetic wastewater.

The synthetic wastewater fed to SBRs included two parts (34). Solution A of 150 ml and solution B of 850 ml were mixed and diluted to 2 liters, where chemical oxygen demand (COD) concentration was 400 mg/liter, PO₄³⁻-P concentration was 20 mg/liter, and the ratio of carbon to phosphorus (C/P) was 20. From the 90th day, COD in feed was up to 600 mg/liter and C/P was up to 30 by adding extra sodium propionate.

Solution A contained the following per liter: 5.84 g (8.76 g from the 90th day) CH₃CH₂COONa, 1.02 g NH₄Cl, 0.01 g peptone, 0.01 g yeast extract, 1.2 g MgSO₄·7H₂O, 0.19 g CaCl₂2H₂O, 7.94 mg allyl-N thiourea (ATU), and 4.00 ml of trace metals solution. Solution B contained 132 mg K₂HPO₄ and 103 mg KH₂PO₄ per liter. The trace metals solution per liter contained 1.5 g FeCl₃6H₂O, 0.15 g H₃BO₃, 0.03 g CuSO₄5H₂O, 0.18 g KI, 0.12 g MnCl₂4H₂O, 0.06 g Na₂MoO₄2H₂O, 0.12 g ZnSO₄7H₂O, 0.15 g CoCl₂6H₂O, and 10 g EDTA.

Chemical analysis.

COD, ammonia (NH₄⁺-N), nitrite (NO₂⁻-N), nitrate (NO₃⁻-N), phosphate (PO₄³⁻-P), and the mixed liquid suspended solids (MLSS) were measured according to standard methods (52). DO and pH were monitored online with DO/pH meters (Multi 340i; WTW, Germany).

Poly-β-hydroxyalkanoates (PHAs) determined by poly-β-hydroxybutyrate (PHB) and poly-β-hydroxyvalerate (PHV) were analyzed (53). FISH probes contain EUB_MIX labeled with fluorescein isothiocyanate (FITC) at the 5ʹ end, including EUB338 (5ʹ-GCT GCC TCC CGT AGG AGT-3ʹ), EUB338II (5ʹ-GCA GCC ACC CGT AGG TGT-3ʹ), and EUB338III (5ʹ-GCT GCC ACC CGT AGG TGT-3ʹ), for all bacteria (54) and PAO_MIX labeled with Cy5 at the 5ʹ end, including PAO462 (5ʹ-CCG TCA TCT ACW CAG GGT ATT AAC-3ʹ), PAO651 (5ʹ-CCC TCT GCC AAA CTC CAG-3ʹ), and PAO846 (5ʹ-GTT AGC TAC GGC ACT AAA AGG-3ʹ), for PAOs (55).

Cell preparation and FISH-staining-flow cytometry.

Sludge samples taken from the 110th day were fixed in 4% polyoxymethylene for 2 h under 4°C, then washed twice with phosphate-buffered saline (PBS) (8.02 × 10³ mg/liter NaCl, 2 × 10² mg/liter KCl, 6.1 × 10² mg/liter Na₂HPO₄, 2 × 10² mg/liter KH₂PO₄), and finally conserved under −20°C. Fixed samples were sonicated in an ice water bath (ultrasonic processor; Sonics and Materials, Inc., USA) for 6 min to approximately 10⁸ cells/ml. After centrifugation to discard supernatant and drying for 10 min, samples were dehydrated for 3 min in a concentration gradient of ethanol (50%, 80%, and 98%) and then resuspended in hybridization buffer containing 5.27 × 10⁴ mg/liter NaCl, 2.42 × 10³ mg/liter Tris-HCl, 0.01% sodium dodecyl sulfate (SDS), 35% formamide (FA), and PAO_MIX probes, followed by incubation for 140 min at 46°C in a dark room. Eventually, samples were washed twice and resuspended in 1 ml PBS. Prior to measure with FCM, samples were filtered through a gauze filter (pore size, 10 μm) and stained for 15 min by adding 10 μl 100× SYBR green I at room temperature.

Double marked cells were detected through FCM (Accuri C6; BD Biosciences, USA). SYBR green I was used to distinguish double-stranded DNA from background noise in FL 1 (49). Cy5 conjugated with PAO_MIX probes was detected in FL 4. The detection threshold was FL 1 = 800, the flow rate was slow (11 μl/min), and 30,000 events were collected. Before the detection, samples were 1,000× or 10,000× diluted to 10⁴∼10⁵ cells/ml. Data were processed using FlowJo v10 to gate cell groups.

DNA extraction and real-time qPCR.

Genomic DNA from freeze-dried activated sludge was extracted using FastDNA spin kits for soil (Bio 101, Vista, CA, USA). The copy numbers of 16S rRNA genes of “Candidatus Accumulibacter” and all bacteria and ppk1 genes of seven “Candidatus Accumulibacter” clades, i.e., clades IA, IIA, IIB, IIC, IIC (ppk1 excluding OTU NS D3), IID, and IIF, were quantified by qPCR (Mx3005P; Agilent Technologies, USA). Specific primers and related qPCR programs are shown in Table 5. The volume of the reaction PCR mixture was 25 μl, using the SYBR Premix Ex Taq GC kit (TaKaRa, Japan), containing 12.5 μl SYBR Premix Ex Taq buffer (2-fold), 8 μl nuclease-free water, 1 μl forward primer (10 mM), 1 μl reverse primer (10 mM), 0.5 μl ROX reference dye II (50-fold), and 2 μl DNA template (10²∼10⁹ copies/μl). The amplification efficiencies of standard curves were in a range of 90 to 120%, and the correlation coefficients were higher than 0.98 (see Table S1 in the supplemental material).

TABLE 5.

Specific primer sequences and programs of qPCR in this study

Primer	Sequence	Target	Amplification size (bp)	Program	Reference
1055F	ATGGCTGTCGTCAGCT	Bacterial 16S rRNA genes	323	95°C for 10 min; 45 cycles of 30 s at 95°C, 60 s at 50°C, 20 s at 72°C	58
1392R	ACGGGCGGTGTGTAC
518f	CCAGCAGCCGCGGTAAT	“Candidatus Accumulibacter” 16S rRNA genes	351	95°C for 3 min; 35 cycles of 30 s at 94°C, 45 s at 60°C, 30 s at 72°C	9
PAO-846r	GTTAGCTACGGCACTAAAAGG
Acc-ppk1-763f	GACGAAGAAGCGGTCAAG	Acc-IA	408	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 61°C, 30 s at 72°C	9
Acc-ppk1-1170r	AACGGTCATCTTGATGGC
Acc-ppk1-893f	AGTTCAATCTCACCGAGAGC	Acc-IIA	105	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 61°C, 30 s at 72°C	9
Acc-ppk1-997r	GGAACTTCAGGTCGTTGC
Acc-ppk1-870f	GATGACCCAGTTCCTGCTCG	Acc-IIB	133	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 61°C, 30 s at 72°C	9
Acc-ppk1-1002r	CGGCACGAACTTCAGATCG
Acc-ppk1-254f	TCACCACCGACGGCAAGAC	Acc-IIC	207	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 66°C, 30 s at 72°C	9
Acc-ppk1-460r	CCGGCATGACTTCGCGGAAG
Acc-ppk1-1123f	GAACAGTCCGCCAACGACC	Acc-IIC (ppk1 excluding OTU NS D3)	254	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 63°C, 30 s at 72°C	9
Acc-ppk1-1376r	ACGATCATCAGCATCTTGGC
Acc-ppk1-375f	GGGTATCCGTTTCCTCAAGCG	Acc-IID	148	95°C for 3 min; 45 cycles of 30 s at 94°C, 45 s at 63°C, 30 s at 72°C	9
Acc-ppk1-522R	GAGGCTCTTGTTGAGTACACGC
Acc-ppk1-355f	CGAACTCGGCGAAAGCGAGTA	Acc-IIF	246	95°C for 3 min; 35 cycles of 30 s at 94°C, 45 s at 70°C, 30 s at 72°C	47
Acc-ppk1-600R	ATCGCCTCCGAGCAACTGTTC

Statistical analyses.

Due to the influence of some unavoidable factors (random error) of FCM, two quantifications of one sample cannot be exactly identical. An equivalence test was carried out to verify the reliability of the FCM experimental data (56). In this study, a more strict tolerance value of 10% was chosen as the criterion for the equivalence test, and the tolerance interval was [−0.1, 0.1] (57). The FCM machine was tested by quality control experiments to ensure that the coefficient of variance (CV) value of fluorescence channels was below 5%.

To assess the relationship between “Candidatus Accumulibacter” clade abundance and physicochemical parameters, ordination biplot of redundancy analysis (RDA) was performed in Canoco 5 to display the distributions of the “Candidatus Accumulibacter” clades with environmental variables. Student’s t tests were carried out in SPSS 19 for the significance of gene abundances determined by qPCR and FCM.