Comparison between two anammox fiber fillers under load impact and the effect of HCO3− concentration

Based on the establishment of a stable anaerobic ammonia oxidation treatment system in 100 days, the impact resistances of two different anammox fiber fillers (the curtain filler: R1 and the bundle filler: BR) were compared. Furthermore, the effect of HCO3− concentration on the bundle filler system was also investigated, the results have shown that the activity of the two anammox fiber fillers was not inhibited when the NO2−–N concentration was lower than 750 mg L−1 (FNA = 0.085 mg L−1), while it was significantly suppressed at 900 mg L−1 (FNA = 0.118 mg L−1). However, the two fiber fillers could be recovered and exhibit a good impact resistance reduction of the substrate concentration. On day 95, the structure of the bundle filler was more conducive to the stable attachment, proliferation, and aggregation of anammox bacteria. Dominant anammox bacteria in both the curtain and bundle fillers were Candidatus Kuenenia, which accounted for 25.9% and 35.9% of the total population, respectively. When the influent HCO3− concentration was 900 mg L−1, the bundled fiber filler had the highest total nitrogen (TN) removal efficiency, which reached 89.0%. Even though it was inhibited under 2000 mg L−1 of HCO3− concentration, the reactor was able to recover within one week by reducing the substrate concentration. In addition, the HCO3− inhibition mechanism was independent of pH, which resulted in high FA content.


Introduction
The anaerobic ammonia oxidation (anammox) process has a high nitrogen removal rate, a low sludge yield, and no need for additional organic carbon requirements. 1 However, the doubling time of anammox bacteria is about two weeks, and the lack of biomass limits further application worldwide. 2 To maintain a high anammox biomass concentration in the reactor, the formation of granular sludge or biolm was found to be a feasible and effective method. The granular sludge reactor has higher volatile suspended matter and its nitrogen removal efficiency is higher than that of the biolm reactor. However, the cultivation of granular sludge requires a long time. Even during stable operation, the granular sludge may be unsteady, and the reactor can show a oating sludge problem. [3][4][5] So far, there is still a controversy about the composition of biolms in general. Among these, the most consistent view is that biolms are composed of loosely distributed amorphous microbial communities under microbial aggregates, including adsorbed substrates and inorganic particles with extracellular polymers surrounding them. [6][7][8] The biolm provides a large specic surface area, strong adsorption capacity, and adequate environment suitable for the enrichment of anammox bacteria. 9 Due to the compact nature of the microbial biolm surface, the anammox process has signicant advantages. 10 Many studies have investigated the substrate removal capacity and retention ability of different anammox biolms made of volcanic rock, zeolite, ceramics, and plastic. 11 However, few studies have focused on the comparison between different types of anammox biolms.
Autotrophic bacteria use inorganic carbon as their carbon source (HCO 3 À ). 12 Sufficient inorganic carbon is needed to enrich anammox bacteria and retain their activity. 13 It has been reported that the activity of anammox bacteria increased as the inuent inorganic carbon concentration increased from 1.0 g L À1 to 1.5 g L À1 , but was inhibited at 2.0 g L À1 . 14 Additionally, excessive concentrations of ammonia form FA (free ammonia) in water, and the pH leads to an increase in FA in the environment. 15 Therefore, the inhibition of the activity of anammox bacteria caused by high HCO 3 À concentrations and FA is possible in nitrogen-containing wastewater. However, in the biolm system, whether the HCO 3 À concentration aggravates the inhibition of FA under high inuent NH 4 + -N needs to be studied urgently. Therefore, in this study, two different ber llers were selected to study the effect of substrate concentrations on the anammox biolm. First, the impact resistance and recovery ability of the two ber llers were compared. The characteristics and community structure of the biolms were also analyzed to provide information for practical applications. Moreover, the effect of HCO 3 À concentration on the nitrogen removal rate and whether the inhibition of HCO 3 À concentration is related to pH and FA were explored.

Materials and methods
2.1. Reactor conguration and operational strategy 2.1.1 Impact load resistance and recovery test. The experiment devices are shown in Fig. 1. Two same up-ow biolm reactors with an effective volume of 20 L were operated for 100 days. There were two partitions inside the reactor: the reaction zone and separation zone. The inoculated sludge was taken from the reserve occulent anammox sludge in the laboratory, and the total nitrogen removal efficiency was beyond 80%. The initial inoculated sludge concentration was about 3800 mg L À1 in two reactors. Additionally, the initial NH 4 + -N and NO 2 À -N concentrations were set as 150 mg L À1 and 180 mg L À1 respectively. The substrate concentration was increased for 9 times in 100 days. For the rst four times, NH 4 + -N and NO 2 À -N concentrations were increased by 100 mg L À1 and 120 mg L À1 , and for the last ve times, they were increased by 50 mg L À1 and 60 mg L À1 . Each stage lasted for 10 d. The pH value of inuent water was controlled at 7.2 AE 0.2. The HRT was set as 6 h. The temperature was controlled at 32 AE 2 C by the heat pipe heating equipment. To maintain anaerobic conditions and prevent the growth of phototrophic organisms, the reactor was wrapped with a black opaque cloth. 16 Although a concentration of NH 4 + -N lower than 1000 mg L À1 cannot inhibit the anammox process, 17 anammox bacteria were sensitive to the NO 2 À -N concentration. 18 Therefore, the NO 2 À -N concentration in the effluent was used to evaluate the activity of anammox bacteria by controlling the temperature and pH. 2.1.2 Inuence of the concentration of HCO 3 À . The bundle ller reactor was selected to study the effect of HCO 3concentration on the TN removal rate, and NaHCO 3 was added as the only source of HCO 3 À for 70 d. Other operating parameters were the same as mentioned in Section 2.1. At the same time, the total nitrogen concentration in the inuent was set to 1320 mg L À1 , and the NaHCO 3 concentration in the inuent was increased in a gradient, as shown in Table 1.

Fiber ller
The xed curtain ller and xed beam ller in the experiment are shown in Fig. 2. Both llers were made of ber. The curtain ller (Jingyuan Environmental Technology Co., Ltd China) was made of acrylic ber and polyester ber. It could use a large amount of biological mass attached to its surface for oxygenation and repeated contact with sewage. Therefore, it could intercept tiny, suspended solids and organic matter that were not easy to precipitate or remove. The bundled ller (Jingyuan Environmental Technology Co., Ltd China) was composed of acrylic ber, polyester ber and reinforcing wire. The ller had a good biolm formation, compact structure with no deformation, and strong resistance to water ow impact. Both llers were suitable for the treatment of high nitrogen-containing wastewater.

Synthetic wastewater
NH 4 Cl and NaNO 2 were used as the nitrogen source of the inuent, for which NH 4 + -N : NO 2 À -N was 1 : 1.2. The other compositions of synthetic wastewater are as follows: NaHCO 3 1 g L À1 ; KH 2 PO 4 0.01 g L À1 ; MgSO 4 0.3 g L À1 ; and CaCl 2 0.056 g L À1 . 1 mL L À1 of trace element concentrate I and II were added, according to Table 2. Nitrogen gas (99%) was purged into the reactor for 20 min to reduce the dissolved oxygen (DO) concentration till it was lower than 0.5 mg L À1 .

À
The batch test was carried out with a series of serum bottles to investigate HCO 3 inhibition, which is independent of pH. The  The serum bottles were sparged with high-purity nitrogen for 20 min to form a dissolved oxygen-free environment. Then, the bottles were sealed and placed in an incubator at a constant temperature of 32 C under dark conditions. 20 The speed was set as 140 rpm. On day 6, day 8, and day 10 aer stable operation, each sample was collected 6 times, each sample was collected 6 times, once an hour, and the average removal rate of TN was calculated to determine the maximum TN removal rate.

Chemical and physical analysis
NH 4 + -N and NO 2 À -N were estimated according to the standard methods. 21 The methods of measurement and instruments are given in Table 3. The above-mentioned samples were measured in duplicate to guarantee the accuracy. The morphology of ber llers was observed using an E400 light microscope (Nikon Corporation, Japan).

DNA extraction and illumina high-throughput sequencing
To compare the differences in microbial diversity between the two ber ller systems, an appropriate amount of biolm was taken from the reactors (the sampling port was located at 20 cm from the bottom of the reactor) on day 95. The collected samples were subjected to high-throughput analysis (curtain ller: R1; bundle ller: BR1). Total community genomic DNA extraction was performed using an E. Z. N. A.Soil DNA Kit (Omega, USA), following the manufacturer's instructions. To ensure that adequate amounts of high-quality genomic DNA were extracted, the concentration of DNA was measured using Qubit 2.0 (life, USA). The V3-V4 hypervariable region of the 16S rRNA gene was amplied using standard protocols. Sequencing was performed using the Illumina MiSeq system (Illumina MiSeq, USA) by Sangon BioTech Company (Shanghai China).

Calculation formula of FA and FNA
The calculation formula of FA and FNA (free nitrous acid) concentrations are as follows: 22 C FA is the concentration of free ammonia, mg L À1 ; C FNA is the concentration of FNA, mg L À1 ; C t,NH 3 is the concentration of total ammonia, mg L À1 ; C t,NO 2 À is the concentration of total nitrite, mg L À1 ; T is the temperature in the system, C. The pH value was controlled at 8 AE 0.2, and the temperature was controlled at 32 AE 2 C.

Nitrogen removal efficiency under different substrate concentrations
With the increased NH 4 + -N and NO 2 À -N concentrations, the nitrogen removal efficiency of the two llers increased at rst and then decreased (Fig. 3). Until the substrate concentration increased to 990 mg L À1 , NO 2 À -N in the system was almost exhausted and the removal rate of NO 2 À -N was stabilized at 99%. When the inuent TN concentration reached 1320 mg L À1 and the FNA reached 0.085 mg L À1 , the nitrogen removal performance of the two llers slightly decreased to about 96% but not inhibited. In the early stage, it could be found that the recovery rate of the bundle packing was better than that of the curtain packing. While in the other sludge system, at this substrate concentration, nitrogen removal rates decreased 90% within 7 d. 23 Moreover, different from slowly deteriorated nitrogen removal efficiency, many studies found that the Glass electrode method ZD-2 automatic potentiometer T Thermometer substrate had caused severe inhibition of anammox activity, and the nitrogen removal efficiency deteriorated signicantly. 24,25 Since high substrate concentration would inevitably bring a high level of FA and FNA, which would inhibit the nitrogen removal efficiency. However, there was no obvious inhibition observed in the reactor, and it suggested that the anammox ber lls have a high tolerance at a concentration of 1320 mg L À1 . When the total nitrogen concentration gradually increased to 1540 mg L À1 , the nitrogen removal efficiency of the bundle ller was better than that of the curtain ller, and the removal rates of NH 4 + -N and NO 2 À -N were 87% and 89%, which indicated that the bundle ller had better resistance to high NO 2 À -N concentrations. On day 90, when the concentration increased to 1650 mg L À1 , the two reactors were inhibited and gradually deteriorated. The removal rates of NH 4 + -N and NO 2 À -N were both less than 20% and the effluent NO 2 À -N concentration was more than 700 mg L À1 . Nitrite is a vital factor affecting the activity of anammox bacteria. 26 Excessive nitrite would form FNA in the liquid, which directly affected the crucial enzymes of microbial metabolism and stimulated the accumulation of intermediate products, thereby inhibiting the metabolism of anammox bacteria. 27 In this study, when the NO 2 À -N concentration was lower than 750 mg L À1 (FNA ¼ 0.085 mg L À1 ), the activity of the both the anammox ber llers was not inhibited. Comparing these results with other studies, 28-30 the threshold value of NO 2 À -N in this reactor was higher. Meanwhile, the FNA threshold value of granular sludge and gel carrier was higher than that of ocs. It was speculated that the biolm thickness or particle size plays an important role in NO 2 À -N threshold value. When the inuent NO 2 À -N concentration reached 900 mg L À1 (FNA ¼ 0.118 mg L À1 ) for 12 h, the activity of anammox bacteria in the reactor decreased signicantly, and foam appeared in the top area of the reactor. These phenomena could indicate that some anammox bacteria died under high NO 2 À -N concentrations. 21 This study showed that the anammox ber lls provided suitable culture conditions for the growth and reaction of anammox bacteria and expanded the resistance range. Unfortunately, the high concentration of nitrite toxicity was still inevitable.

Comparison of the impact resistance between two ber llers
The two ber ller systems stably operated when the inuent TN concentration was 770 mg L À1 (Fig. 4). It proved that even though the nitrogen removal efficiency decreased, the ber ller system still had high activity aer recovery by a decreased substrate concentration. Aer the inuent TN concentration increased to 1430 mg L À1 in the rst 12 h, there was no signicant difference in the removal efficiency between the two kinds of llers. The NH 4 + -N concentration in the effluent was 44.0 mg L À1 and 35.9 mg L À1 , and the NO 2 À -N concentration was 25.8 mg L À1 and 25.2 mg L À1 , respectively. However, in the last 12 h of the cycle, the nitrogen removal efficiency of the two ber llers showed an obvious difference. The nitrogen efficiency of the bundle ller was more stable than that of the curtain ller, and the difference in total nitrogen concentration of the effluent was more than 70 mg L À1 . In the subsequent recovery stage, the two ber ller systems recovered rapidly within 24 h aer the inuent matrix concentration restored to 770 mg L À1 . The NH 4 + -N/NO 2 À -N concentrations of the effluent of the curtain ller and bundle ller were 22.39/14.19 mg L À1 and 17.44/13.07 mg L À1 , respectively. Within 1 d aer recovery under low inuent substrate concentrations, the NH 4 + -N concentration in the effluent of the bundle ller almost recovered to the value before the impact test, and the effluent NO 2 À -N concentration also decreased signicantly. Aer only 3 d, the effluent NH 4 + -N and NO 2 À -N concentrations were reduced to below 15 mg L À1 and 11 mg L À1 , respectively. The recovery of  the bundle ller was faster than that of the curtain ller during the increase in substrate concentration.

Comparison of ber llers before and aer biolm formation
The two biolm llers were observed before and aer the bio-lm formation using an E400 light microscope (Nikon Corporation, Japan) on day 95. It could be seen from Fig. 5 that the surface of the llers was clean and smooth before the anammox biolm was hanging lmed. The ber material (Fig. 5(a)) had a small and dense packing structure, which was helpful for trapping anammox bacteria, while the reinforced ber material (Fig. 5(b)) had a smooth surface under the microscope, which was more conducive to supporting the brous packing. Fig. 5(c) and (d) show the microstructures of ber biolms. The entire biolm was reddish-brown, which was consistent with the characteristics of anammox bacteria. Fig. 5(e) shows a kind of reinforcing ber, with a small number of bacteria distributed on the surface, which proved that it also had some retention ability but was weaker than that of the ber ller. 31 Due to the adhesion and accumulation of bacteria on the ber surface, the weight of the ber increased. However, the reinforcing ber was easy to be washed with water, resulting in a loose ber structure. The curtain ller had a strong hydrophilic ability and large specic surface but was vulnerable to the impact of the inuent and oating with the water ow uctuation. Aer most bacteria adhered to the ber surface, the weight of the ber increased, and the ber was easy to fracture with the water ow, and new llers needed to replace regularly. The bundled ller was made of dense bers with high toughness and had a large specic surface area. Therefore, many bacteria were attached to the bundled biolm, and the ber was not easy to fall off. Meanwhile, it had a certain tolerance to the erosion of water. In conclusion, the curtain ller and bundle ller were different in structure and application performance. Due to its meticulous structure, the bundle ller increased the specic surface area, which was conducive to improving the attachment of anammox bacteria. Moreover, laments strengthened the toughness of the bundle ller, which could resist the scouring force of water ow and prevent the bacteria and bers from falling off easily.

Microbial community structure and abundance of two llers
The phyla of curtain ller and bundle ller were compared (Fig. 6). Proteobacteria accounted for 18.3% of the total bacterial community in the bundled ller, which was 5.3% lower than that of the curtain ller. Planctomycetes (the phylum of anammox bacteria) in the bundle packing and curtain packing accounted for 41.1% and 35.7% respectively, which were higher than 15.8% in other studies. 32 These results may explain that the anammox biolm system was better than the activated sludge system with a high nitrogen concentration. Compared with the curtain ller, the bundle ller was more conducive to the growth and attachment of Planctomycetes. In addition, the difference in Bacteroides between the two ller samples was 2.1%. Bacteroidetes were chemoheterotrophic bacteria. Although no exogenous organic substances were added into the anammox reactor, the metabolic growth and reproduction of organisms in the systems would inevitably produce organic substances, increasing Bacteroidetes. Moreover, the bacteria  belonging to Chloroexi and Bacteroidetes have some impact on sludge granulation.
Combined with Section 3.3, it could be seen from Fig. 7 that Candidatus Kuenenia, Candidatus Brocadia, and Candidatus Anammoxoglobus accounted for a high proportion in both curtain llers and bundle llers, and Candidatus Kuenenia dominated in the system, which gives an advantage for the anammox process. Candidatus Kuenenia could utilize low NO 2 À -N concentration and has good tolerance to high NO 2 À -N concentrations, which is consistent with the results of this study. 33 The results of high-throughput sequencing also conrmed that the content of Candidatus Kuenenia in the bundle ller was higher than that in the curtain ller. The number and distribution of anammox bacteria on the surface of the curtain and bundle llers were increased and evenly distributed, and the biolm tended to mature. It was proved that the appropriate substrate concentration in the environment was conducive to the proliferation and aggregation of anammox bacteria. Moreover, it conrmed that the two kinds of llers were suitable for nitrogen removal in wastewater, but bundle ller was better.

Effect of HCO 3 À concentration on nitrogen removal performance
To explore the optimum and inhibition value of HCO 3 À concentration in the ber ller system, the HCO 3 À concentration gradually increased when the inuent TN stabilized at 1320 mg L À1 . It could be seen from Fig. 8  concentration was further increased to 2000 mg L À1 , the TN removal rate dropped to 50% within 5 d, and the effluent TN concentration increased to 650 mg L À1 , which was lower than the inhibition level (75%) in other studies. 36 However, once the HCO 3 À concentration was reduced to 900 mg L À1 on day 64, the TN removal rate of the bundled ber ller recovered to 85% in a short time, and the effluent NO 2 À -N concentration was lower than 20 mg L À1 . It also reported similar bicarbonate inhibition of short recovery period. 37 A hypothetical explanation proposed for this phenomenon was that the inhibited effect was due to high FA concentrations. However, HCO 3 À concentration less than 1000 mg L À1 may cause the lack of inorganic carbon, further inhibiting the anammox process. For the ber ller system, due to the lower HCO 3 À consumption and the uniform distribution of bacteria, the HCO 3 À inhibition concentration was lower. On the contrary, since anammox bacteria were in the core of the ocs, they could act better against the increasing HCO 3 À concentration. Low HCO 3 À concentrations could lead to a decrease in pH, resulting in the overow of carbon dioxide during the aeration phase of the SBR. However, at a HCO 3 À concentration of 1000 mg L À1 , no bicarbonate limitation was observed in the reactor. The stoichiometric ratio could reect the stability of the anammox process. As shown in Fig. 9 37 It was found that the optimal HCO 3 À concentration was 1500 mg L À1 , and the ratio of NH 4 + removed to NO 2 À removed was   (900 mg L À1 ), the stoichiometric ratio ranged between 0.31 and 0.42, exceeding the theoretical value of 0.26. According to the anammox reaction process, every 1 mol of NH 4 + -N will consume 0.13 mol of H + and 0.066 mol of HCO 3 À , which increased the pH of the effluent, 39 but HCO 3 À was used as the pH value. The amount of buffering agent would also affect the pH of the effluent. The optimal pH value of the anammox reaction was in the range of 6.7-8.3, 40 and the removal of excess nitrogen load would cause the pH value in the reactor to increase seriously, thereby inhibiting the anammox reaction. It can be seen from Fig. 8 that when the inuent HCO 3 À concentration was less than 900 mg L À1 , the effluent pH value was greater than 8.3, which was not suitable for the growth of anammox bacteria. When the inuent HCO 3 À concentration was greater than 900 mg L À1 , the pH of the effluent could be stabilized in the optimal range of pH required for the anammox reaction. The difference in pH between the inuent and the effluent was minimal, which was more conducive to the progress of the anammox reaction.
3.6. HCO 3 À inhibition pathway on the anammox bundle As shown in Fig. 10, as the HCO 3 À concentration increased from 100 mg L À1 to 800 mg L À1 , the removal rate of TN increased from 105 mg N (L h) À1 to 188 mgN (L h) À1 . It could be explained that when the HCO 3 À concentration was lower than the optimal concentration, the lack of inorganic carbon resulted in a lower TN removal rate. When the inuent HCO 3 À concentration was too high, although the pH remained at 7.2 (AE0.2), and the FA concentration was lower than 5 mg L À1 , it still had a negative effect on the anammox process. 41 Therefore, it can be concluded that inorganic carbon inhibition was independent of the effect of bicarbonate on pH; therefore, it was not caused by an increase in FA. Compared with the experiment conducted under continuous conditions, the optimal HCO 3 À concentration detected in the batch experiment was slightly shied, which may be due to the increase in TN load. Rather than the constant load rate used in batch tests. Under the over-ow conditions of the ller reactor, partial FA suppression also contributed to the overall suppression, and bicarbonate was most likely to be the main contributor.

Conclusions
The anammox activity of two anammox ber llers is favorable. When the inuent TN concentration reached 1320 mg L À1 , two biolm processes were not inhibited, and they still had high activity. Even with a concentration above 1650 mg L À1 , the two kinds of ller reactors had a good impact resistance aer substrate reduction, and the bundle ller showed better performance. Moreover, the curtain ber and bundle llers had different ways of interception of anammox bacteria. The curtain packing was highly hydrophilic, but the packing was susceptible to the impact of ingress of water. However, the bundle ller was strong with toughness, and the bers were not easy to fall off, which was more conducive to the proliferation and aggregation of anammox bacteria. High-throughput sequencing revealed Candidatus Kuenenia was the dominant genus in the two bio-lms, and a higher content was observed in the bundle ller (35.9%) than in the curtain ller (25.9%), which showed better efficiency and tolerance to high NO 2 À -N concentrations. When the inuent HCO 3 À concentration was 900 mg L À1 , the bundled ber ller had the highest removal efficiency. The inhibition can be quickly released by reducing the HCO 3 À concentration.
In addition, the HCO 3 À inhibition mechanism was independent of pH, which resulted in high FA content.

Conflicts of interest
The authors of this manuscript have no conicts to declare.