Diversity and abundance of denitrifying bacteria in a simultaneously nitrifying and denitrifying rotating biological contactor treating real wastewater at low temperatures

Nitri ﬁ cation and denitri ﬁ cation processes occur simultaneously in aerobic wastewater bio ﬁ lms. Although wide regions of the world have average temperatures of less than 15 °C for a half year, few studies have investigated the nitrogen removal by nitri ﬁ cation and denitri ﬁ cation in a single-stage aerobic bio ﬁ lm reactor used for treating real wastewater under low-temperature conditions. This study showed successful wastewater treatment in a high average nitrogen removal rate of 78% at low water temperatures by simultaneous nitri ﬁ cation and denitri-ﬁ cation in a rotating biological contactor (RBC) bio ﬁ lm. Batch operations using the RBC to evaluate the rates of ammonium decrease at low temperatures demonstrated that the rate of ammonium decrease at 8 °C was 76% of that at 20 °C. Daily monitoring of nitri ﬁ cation and denitri ﬁ cation rates suggested that the denitri ﬁ cation rate was highly correlated with the nitri ﬁ cation rate. Next-generation sequencing (NGS) analysis revealed the presence of diverse and abundant denitrifying bacteria and aerobic bacteria in the RBC bio ﬁ lm more than those in the activated sludge samples, which probably enabled the achievement of the high nitrogen removal rates at such low temperatures. Furthermore, correlation with the colony counts showed that the NGS analysis had the quanti ﬁ cation range of three orders of magnitude (from 0.001% to 1%).


INTRODUCTION
Simultaneous nitrification and denitrification (SND) in a wastewater treatment process has been studied over the past four decades and has increasingly become a focus of attention in the last 10 This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). years because of the worldwide requirement for high water quality of discharged wastewater, at a smaller footprint and with less energy consumption. To achieve denitrification under aerobic conditions, it is important to carefully control the concentration of dissolved oxygen (DO) at a low level, especially in the case of suspended growth sludge such as in oxidation ditch systems (Metcalf & Eddy 2003). On the contrary, in the case of attached growth sludge such as biofilms, anaerobic zones are created inside the biofilm, even under submerged oxic or non-submerged conditions (Kühl & Jørgensen 1992;Okabe et al. 1999a). This is because oxygen in the surface layer of the biofilm is consumed by aerobic organisms, so that both nitrification and denitrification occur within the biofilm (Masuda et al. 1991;Satoh et al. 2003).
One of the biofilm-based wastewater treatment technologies is the rotating biological contactor (RBC). A partially submerged (typically 40%) RBC is effective in enhancing SND reaction because the RBC biofilm is periodically exposed to air and underwater conditions, so that nitrification and denitrification occur in turn, for instance, every 10 s, when the rotating speed is set at the optimum of 3 rpm (Watanabe et al. 1992). Thus, we believe that the RBC biofilm has great potential for achieving stably high performance of SND.
Most of SND studies have been conducted at temperatures of 25°C, 30°C, or higher. However, wide regions of the world (e.g., Europe, Central Asia, Eastern Asia, and Northern America) have average temperatures of less than 15°C for a half year and less than 10°C for a period of 3 months (Japan Meteorological Agency 2018). Only a few SND studies have been performed at such low temperatures (de Kreuk et al. 2005;Luostarinen et al. 2006). Moreover, there has been no study using RBC that focused on SND for nitrogen removal from real municipal wastewater at low temperatures. However, real wastewater at low temperatures should significantly affect the diversity of the key microbial groups of the RBC biofilm.
Nitrification and nitrifying bacteria have been studied in the RBC biofilm by cultivation-based techniques, 16S ribosomal RNA (rRNA)-based approaches and using microelectrodes. Since nitrifying bacteria are phylogenetically clustered in only several lineages of a 16S rRNA phylogenetic tree (Purkhold et al. 2003), they have been successfully identified by a fluorescent in situ hybridization (FISH) technique (Juretschko et al. 1998;Okabe et al. 1999b). In contrast, FISH and conventional cloning techniques have been impractical for the detection of denitrifying bacteria, which are phylogenetically spread over very diverse lineages (Graf et al. 2014). Therefore, information on the abundance and diversity of denitrifying bacteria involved in SND in the RBC biofilm is very limited. Currently, next-generation sequencing (NGS) allows analysis of a 100 or 1,000 times higher number of sequences than that analyzed by conventional cloning, which enables the comprehensive detection and analysis of relative abundance of denitrifying bacteria among total bacteria. This study conducted a 2-month continuous operation of an aerobic RBC for real wastewater treatment and examined the performance of simultaneous nitrogen removal at low temperatures. Furthermore, to understand characteristics of the RBC biofilm, the diversity and relative abundance of denitrifying bacteria and nitrate-utilizing bacteria in the RBC biofilm were investigated based on NGS analysis of the microbial community and compared with those in several activated sludges from real wastewater treatment plants (WWTPs). Finally, the quantification range and resolution of the NGS analysis were assessed by comparing with the sequence read frequency with the Bacillus colony count.

Rotating biological contactor
A bench-scale rotating-disk reactor consisted of five stainless steel mesh disks with a diameter of 0.6 m. The reactor volume was 73 L, and the hydraulic retention times (HRTs) were 8 h (until day 22) and 12 h (day 23-31), corresponding to the hydraulic loading rates of 3.23 and 2.15 L/m 2 /h, respectively. The disk rotational speed was 1.5 rpm. The disks were 40% submerged.
The reactor was operated for 2 months, from December 2016 to February 2017. To speed up the start-up process, the reactor was inoculated with 100 g of freeze-dried sludge from the excretion treatment plant (Minakami, Gunma, Japan), which was supplemented with 4 L of milk and minerals and then operated in a batch mode for 2 days. The influent was the toilet wastewater of the Gunma College of Technology. The average influent water quality is described in Table 1. The average influent biochemical oxygen demand (BOD) and NH 4 þ concentration were approximately 300 and 40 mg/L, respectively.

Activated sludge samples
Activated sludge samples were collected from six domestic WWTPs and one pig farm WWTP: Gunma Kenou (GK) WWTP with a capacity of 240,000 m 3 /day, Nagaoka-Chuo (NC) WWTP with a capacity of 68,400 m 3 /day, Gunma Okutone (GO) WWTP with a capacity of 213,000 m 3 /day, Maebashi Omuro (MO) WWTP with a capacity of 756 m 3 /day, Maebashi Baba (MB) WWTP with a capacity of 200 m 3 /day, China Xining (CX) WWTP with a capacity of 85,000 m 3 /day, and Hayashi pig farm (HF) WWTP with a capacity of 180 m 3 /day and BOD of 5,500 mg/L. GK, NC, and GO operate the conventional activated sludge process. MO, MB, CX, and HF operate oxidation ditch, an intermittently aerated activated sludge, an aerated activated sludge after RBC, and an aerated lagoon, respectively.

Measurements of ammonium, nitrite, and nitrate
Ammonium was determined by the Nessler method after sample filtration to remove suspended particles. Similarly, nitrate plus nitrite were routinely determined after sample filtration by an ultraviolet spectrophotometric screening method (APHA, AWWA, WEF 2005). The concentration of each nitrate and nitrite was determined by a naphthyl-ethylenediamine spectrometric method combined with an automated cadmium reduction method (APHA, AWWA, WEF 2005) using an Auto Analyzer (BL Tec K.K., Tokyo, Japan) equipped with a cadmium reduction column. The data confirmed that nitrite was a minor component of ionic nitrogen, and its concentration was always below 1 mg-N/L. The daily nitrification rate (mg-N/m 2 /h) was calculated as follows: (influent NH 4 þ -effluent NH þ 4 )Â reactor volume/RBC surface area/HRT. The daily denitrification rate (mg-N/m 2 /h) was calculated The NH 4 þ decrease rate and NO À x (NO 2 À plus NO 3 À ) increase rate were determined at 20°C, 13°C, and 8°C by batch operation of RBC with the addition of an NH 4 Cl solution to a final concentration of 20 mg-N/L at the beginning of the batch incubation. The concentrations of NH 4 þ and NO x À during the initial 3-h batch incubation were used to calculate the NH 4 þ decrease rate and the NO x À increase rate. The SND efficiency represents the disappeared NO 2 À plus NO 3 À (denitrification) per NO 2 À plus NO 3 À expected to be produced from the decreased NH 4 þ (nitrification). The SND efficiency was calculated by subtracting the NO x À increase rate from the NH 4 þ decrease rate and then dividing the value by the NH 4 þ increase rate using the following equation: Enumeration of Bacillus population One milliliter of an RBC biofilm sample was suspended in 20 mL of a sterile 0.9% (w/v) sodium chloride solution. For an activated sludge sample, 20 mL of a mixed liquor was used for the following procedure. The suspended biofilm sample or activated sludge sample was homogenized and serially diluted by a 10-fold dilution series with a sterile sodium chloride solution. One hundred microliters of a serially diluted sample were plated onto Bacillus agar medium (modified from Murakami et al. 1995), and the plates were incubated at 32°C for 48 h. The Bacillus agar medium contained (per liter): nutrient broth, 8 g; glucose, 8 g; soluble starch, 10 g; sodium chloride, 6 g; agar, 17 g. The typical Bacillus colonies were counted as described by Hatamoto et al. (2017). Some of the colonies were picked and phylogenetically confirmed to belong to the genus Bacillus by 16S rRNA gene sequencing analysis.

16S rRNA gene sequencing
Samples collected from the RBC reactor and seven different activated sludge aeration tanks were used for microbial community analysis. The RBC samples were collected at day 0, day 4, and day 14 after a start-up period of 1 month when the biofilm thickness became constant. Each sample was analyzed as follows and averaged to calculate the relative abundance of denitrifying bacteria. One of the RBC samples at day 14 was sectioned into four layers using cryomicrotome (CM3050S, Leica Microsystems, Inc., Nussloch, Germany). DNA extraction, polymerase chain reaction (PCR), NGS, and quantitative insights into microbial ecology (QIIME) analysis were carried out at the Bioengineering Lab. Co., Ltd (Atsugi, Japan). Genomic DNA was extracted using a FastDNA spin kit for soil (MP Biomedicals, Santa Ana, California, USA) according to the manufacturer's instructions. Then, the V3-V4 region of the bacterial 16S rRNA gene was PCR-amplified using the primers S-D-Bact-0341-b-S-17 and S-D-Bact-0785-a-A-21 (Herlemann et al. 2011) according to the Illumina, Inc., Part #15044223 Rev. B. The PCR products were sequenced using an Illumina MiSeq instrument. The sequence data were analyzed using QIIME (Caporaso et al. 2010). Approximately 50,000 reads of 430-bp long were obtained for each sample. A chimera check and operational taxonomic units (OTUs) clustering were performed, and then OTU representatives were analyzed phylogenetically against the Greengenes database (DeSantis et al. 2006).

Performance of a continuous RBC reactor treating real wastewater at low temperatures
The water quality of the influent and effluent of the continuous RBC reactor was monitored for 2 months at water temperatures ranging from 6 to 18°C during the winter period ( Figure 1). The average water quality of the influent wastewater is summarized in Table 1. After a start-up period of 1 month, approximately 20 mg-N/L ammonium decreased by nitrification, whereas there was no equivalent increase of nitrate and nitrite, and thus 36% of total inorganic nitrogen was lost. The sum of the influent NH 4 þ and NO x À was 44 + 10 mg-N/L, and the sum of the effluent NH 4 þ and NO x À was 28 + 14 mg-N/L. The results indicated that part of the nitrate produced by nitrification was simultaneously denitrified in the RBC biofilm by a so-called SND process. The efficiency of SND was 69 + 15% when operated at HRT of 8 h (until day 22) and 78 + 10% when operated at HRT of 12 h (day 23-31) (Figure 1(c)). In the case of the conventional activated sludge process for aerobic wastewater treatment, the decrease in ammonium was simultaneous with the increase in nitrate by nitrification. Figure 2 shows the correlation between daily nitrification and denitrification rates during the RBC operation. Although both rates fluctuated during the 30-day RBC operation, interestingly, the daily change of the denitrification rate completely coincided with the daily change of the nitrification rate (Figure 2(a)). In addition, the denitrification rates never exceeded the nitrification rates, even when a higher concentration of nitrate than that of ammonium was expected to exist in the bulk water, for example, after the peaks of the daily nitrification rate. The daily nitrification rates were highly correlated with the daily denitrification rates (Figure 2(b)), and the daily denitrification rates were at approximately 80% of the daily nitrification rates, based on the approximate line. These results demonstrated that the denitrification reaction was dependent on the nitrification reaction in the RBC biofilm. In other words, nitrification was a rate-limiting step in the SND.
Ammonium decrease rate and nitrate/nitrite increase rate at low temperatures The ammonium decrease rate and nitrate/nitrite production rate were determined at low temperatures in a batch experiment using ammonium as a sole nitrogen compound. The ammonium decrease rates (i.e., nitrification rates) were 54, 56, and 44 mg-N/m 2 /h at temperatures of 20°C, 13°C, and 8°C, respectively ( Figure 3). Thus, the ammonium decrease rate at 8°C was still 76% of that at 20°C, suggesting that the nitrifying bacteria in the RBC biofilm were tenacious at such low temperatures. The difference between the NH 4 þ decrease rate and the NO x À increase rate reflects the denitrification rate, which is expressed as the SND efficiency in Figure 3. Although the SND efficiency was lower at lower temperatures, that at 8°C was still 29%, i.e., approximately one-third of oxidized ammonium was denitrified in the aerobic RBC biofilm.

Phylogenetic diversity of denitrifying and nitrate-utilizing bacteria
Microbial communities of RBC biofilms were analyzed by NGS. The relative abundances of denitrifying bacteria, including nitrate-utilizing bacteria, in the microbial communities are shown in Figure 4(a) and Table 2. Many of the 16S rRNA gene sequences in the RBC biofilm were associated with denitrifying and nitrate-utilizing bacteria. The most abundant phylotypes were related to the genera Rhodobacter (relative abundance: 3.9%), Zoogloea (relative abundance: 1.9%), and Leucobacter (relative abundance: 1.9%), whose members are capable of denitrification (Weon et al. 2012;Bellini et al. 2013;Graf et al. 2014;Unz 2015). Acidovorax, Hydrogenophaga, and Comamonas (sum of relative abundances: 2.8%), which belong to Comamonadaceae, are well-known abundant phylotypes in activated sludge, many species capable of heterotrophic denitrification (Graf et al.  . Haliscomenobacter (relative abundance: 1.3%), a thin-sheathed bacterium, has been shown to utilize ammonium and nitrate as nitrogen sources (van Veen et al. 1973). The genus Bacillus was one of the most abundant phylotypes. The BLAST analysis revealed that the most closely related sequences were those of Bacillus subtilis and Bacillus tequilensis, with a 100% match (both species have identical sequences in the sequenced region). Although Bacillus species are not well-known denitrifying bacteria, B. tequilensis has been shown to reduce nitrate to nitrogen gas (Gatson et al. 2006). B. subtilis reduces nitrate to nitrite and ammonium during nitrate respiration (Logan & Vos 2015). Some strains of B. subtilis have been reported to perform heterotrophic nitrification and denitrification (Kim et al. 2005;Yang et al. 2011). Thus, the RBC biofilm harbored a higher diversity and abundance of denitrifying bacteria and nitrate-utilizing bacteria.  Phylotypes related to denitrifying and nitrate-utilizing bacteria were significantly less abundant in the activated sludge samples. The relative abundance of total denitrifiers and nitrate utilizers in the RBC biofilm was 21%, whereas that in the activated sludge samples was 6% (Figure 4(a) and 4(c)). Most of the phylotypes affiliated with denitrifiers (14 of 17) were more abundant in the RBC biofilm than in the activated sludge samples ( Table 2). The phylotypes affiliated with the genera Leucobacter, Paracoccus (van Spanning et al. 2015), Thermomonas (McIlroy et al. 2016), Haliscomenobacter, Hydrogenophaga, Thauera (Liu et al. 2013;Heider & Fuchs 2015), Comamonas, and Kaistia ( Jin et al. 2012) were found in the RBC biofilm at an abundance of one or two orders of magnitude higher than that in the activated sludge samples. The relative abundances of the phylotypes affiliated with the genera Rhodobacter, Acidovorax, Haliangium, and Bacillus were five or six times higher in the RBC biofilm than in the activated sludge samples. These results confirmed that the RBC biofilm provided a more favorable habitat for denitrifying bacteria and nitrate-utilizing bacteria than did the activated sludge. The depth profile of the denitrifying and nitrate-utilizing bacteria revealed that these bacteria were distributed throughout the biofilm, from the surface to the bottom, and were more abundant at a depth of 400-700 μm (Figure 4(b)). It seemed that the RBC biofilm was less suitable as a habitat for aerobic bacteria than was the activated sludge. However, the populations of Bacillus, Flavobacterium, and Thermomonas, which are generally recognized as aerobic bacteria, were also distributed throughout the depth of the biofilm. Microorganisms inhabiting the RBC biofilm are exposed to a periodic, 'second-order' change between oxic and anoxic conditions (Watanabe et al. 1992), in contrast to the bacteria living in the activated sludge, where the concentration of DO is relatively stable. The time length of the alternating oxic and anoxic conditions is 20 s when the rotation speed is set at 1.5 rpm. The periodically changing conditions could enhance the abundance and diversity of denitrifying bacteria, nitrate-utilizing bacteria, and aerobic bacteria throughout the RBC biofilm.

Quantification by NGS
Bacillus colonies grown on Bacillus agar medium have morphologically distinctive features, and thus the Bacillus population was quantified by a colony-counting method. Figure 5 illustrates a logarithmic correlation between the Bacillus colony counts and the sequence read frequency of Bacillus determined by NGS. Based on the generated approximate line, 1% relative abundance of Bacillusaffiliated sequences corresponded to 1 Â 10 9 Bacillus colony-forming units (CFU)/g mixed liquor suspended solids (MLSS). The correlation was described by the equation: y ¼ 10 9 Á x 1:00 . The one order of magnitude difference by colony counting was almost the same as the one order of magnitude difference by the NGS detection frequency. Although NGS is a high-throughput technology, DNA extraction and PCR are still unavoidable when analyzing microbial communities, and the bias caused by these steps affects the quantification of the relative abundance of each microbial population. Nevertheless, our NGS analysis had the quantification range of three orders of magnitude, from 0.001% to 1%, with the quantification resolution similar to that obtained by colony counting.
The Bacillus population of the RBC biofilm was, on average, 1.3 + 0.5 Â 10 9 CFU/g suspended solids, while those of activated sludge were one or more orders of magnitude lower and were in the range from 7.2 Â 10 5 to 2.7 Â 10 8 CFU/g MLSS ( Figure 5). Hatamoto et al. (2017) have recently reported that the Bacillus populations of 12 activated sludge samples were in the range between 1.0 Â 10 7 and 6.6 Â 10 8 CFU/g MLSS. Microbial community analysis using NGS also showed that the relative abundance of Bacillus species was the highest in the RBC biofilm (1.2 + 0.6%) and was lower in the activated sludge samples (0.001-0.7%) ( Figure 5). Thus, a higher abundance of Bacillus species in the RBC biofilm compared with that in the activated sludge samples was confirmed by both the colony-counting method and NGS. Mechanism of SND in the RBC biofilm Table 3 summarizes that the RBC reactor increased the efficiency of SND when operated in the continuous flow rather than in the batch mode and that high SND efficiency was obtained at lower ORP (oxidation-reduction potential) values. Although we did not investigate how ORP reached low values in the continuous RBC reactor, we expected that the ORP value was lowered by the high activity of both aerobic bacteria and denitrifying bacteria inhabiting the biofilm at all depths (Figure 4(b)). When influent BOD was continuously fed to the reactor, oxygen consumption by aerobic bacteria probably provided a favorable habitat for denitrifying bacteria in the RBC biofilm. On the other hand, in the batch mode operation, BOD was probably limited for these bacteria. Therefore, insufficient DO was consumed to drop the ORP value below zero, and so, SND in the batch mode operation did not increase to the levels of SND efficiency observed in the continuous flow operation. These observations suggested that the abundance of aerobic bacteria such as Thermomonas, Flavobacterium, and Bacillus species (Figure 4 and Table 2) was a key to achieve high SND efficiency in the RBC biofilm.

CONCLUSION
This study demonstrated highly efficient nitrogen removal by SND in a single-stage aerobic RBC reactor used for real wastewater treatment at low temperatures. The denitrification reaction was highly dependent on the nitrification reaction compared to the concentration of nitrate. NGS analysis revealed that the RBC biofilm harbored a higher diversity and an abundance of denitrifying bacteria, nitrate-utilizing bacteria, and aerobic bacteria than that by the activated sludge. Furthermore, these bacteria were distributed throughout the biofilm, from the surface to the bottom. The periodic, 'second-order' change between oxic and anoxic conditions in the RBC biofilm probably provided a favorable habitat for these bacteria and therefore enabled SND. The correlation between the colony counts and the sequence read frequency determined by NGS revealed that the NGS analysis had the quantification range of three orders of magnitude (from 0.001% to 1%).

ACKNOWLEDGEMENTS
This study was supported by the Japanese MLIT Gesuido Academic Incubation to Advanced (GAIA) project.