Enhanced nitrogen removal via SNAD-IFAS process treating THP-AD liquor: microbial community and co-metabolism mechanism

Performance of the SNAD-IFAS hybrid process

The performance of the SNAD-IFAS hybrid process at different phases is shown in Figs. 1 and 2. The first phase was the start-up of the SNAD-IFAS hybrid process. The influent was the diluted THP-AD liquor, in which NH₄⁺-N and COD concentrations were 445.0 and 715.0 mg/L, respectively. From day 0 to day 12, the ammonium removal efficiency increased from 85.1% to 100% and the total inorganic nitrogen (TIN) removal efficiency increased from 74.4% to 89.9% (Figs. 1b and 2b). During the stable operation of the start-up, the average NH₄⁺-N, NO₂⁻-N, and NO₃⁻-N concentrations were 0, 10.9, and 50.3 mg/L in the effluent of the hybrid process, respectively (Fig. 1a, c, d). COD was removed with an average COD removal efficiency of 20.7%, and the average COD concentration was 355.6 mg/L in the effluent during the whole phase I (Fig. 2c, d). When the TIN removal efficiency of the hybrid process was 89.9%, the TIN removal efficiencies of SIR1 and SIR2 were 63.1% and 23.7%, respectively, indicating that the nitrogen removal of the hybrid system mainly depended on SIR1 and SIR2. The SNAD-IFAS reactors in the hybrid system can be quickly started in 12 days because the inoculated activated sludge and carriers were taken from a full-scale SNAD reactor.

a Influent (Inf.) and effluent (Eff.) concentration of TIN. b TIN removal efficiency. c Influent (Inf.) and effluent (Eff.) concentration of COD. d COD removal efficiency.

In Phase II (13–106 d), the concentration of NH₄⁺-N in the influent was ranged from 736.0 mg/L to 1347.7 mg/L (except for the fluctuation point) by progressively decreasing the dilution ratio, and the C/N ratio of the influent was kept at 1:1. To maintain the nitrogen removal efficiency of SIR1 and SIR2, the values of DO and pH were adjusted to cope with the increase in the NH₄⁺-N concentration in the influent. It was shown that the NH₄⁺-N concentration in the influent of SIR1 was approximately 736.0–1347.7 mg/L, which is significantly higher than that of SIR2 (98.8-553.9 mg/L) (Fig. 1a). Thus, more DO was needed for nitrosation in the SIR1, which can not only provide a substrate for AnAOB but also reduce the inhibitory effect of FA on AnAOB activity, while the SIR2 was controlled under low DO conditions to ensure the activity of AnAOB. A previous study also demonstrated that the DO is usually maintained at about 0.30 mg/L in order to sustain the partial denitrification process. Similarly, to maintain the activity of AnAOB, DO was controlled at around 0.10 mg/L¹⁹. Therefore, the DO concentrations of SIR1 and SIR2 were controlled at 0.15–0.35 mg/L and 0.10–0.20 mg/L, respectively (Table 1). Although the DO of SIR1 was only 0.05–0.15 mg/L higher than that of SIR2, even subtle differences in DO concentration could show different treatment effects due to the extremely strong activity of the AOB enriched in the SNAD-IFAS system (Fig. S1). At the same time, it was also this extremely strong activity of AOB that ensured that a higher partial denitrification process could be achieved even under low DO conditions, providing a reaction substrate for anammox bacteria.

In addition to adjusting the DO, it was also necessary to adjust the pH of SIR1 and SIR2, which could adjust the concentrations of FA in the two reactors to ensure the nitrogen removal efficiency of the two reactors. The reason for controlling the concentration of FA was that different functional bacteria had different tolerance capacities for FA. It was reported that the activity of AOB would be inhibited when the concentration of FA was in the range of 10–120 mg/L, while it was 0.1–1.0 mg/L for NOB^24,25. NOB was more sensitive to FA than AOB²⁵. If the concentration of FA was greater than 30 mg/L, the activity of AnAOB was also inhibited²⁶. Therefore, the pH of SIR1 and SIR2 was controlled at 7.7–7.8 and 8.0–8.1 (adjustment by adding potassium bicarbonate), respectively (Table 1), to ensure that the FA concentrations of SIR1 and SIR2 were about 16-23 mg/L and 3 mg/L, respectively. Through the adjustment of the above two process parameters, on the 48th day, the TIN concentration of the influent reached more than 1000 mg/L, and the TIN removal efficiencies of SIR1 and SIR2 were 76.2% and 14.0%, respectively (Figs. 2a, 3b). The TIN removal efficiency of the SNAD-IFAS hybrid process was 92.3% (Fig. 2b). On the 96th day, SIR1 and SIR2 can already treat the undiluted THP-AD liquor. The average TIN removal efficiencies of SIR1 and SIR2 were 70.6% and 15.4%, respectively (Fig. 2a, b). The average TIN removal efficiency of the SNAD-IFAS hybrid process was 89.9% (Fig. 2b), which was higher than the one-stage and two-stage PN/A process^11,12,13,14. During phase II, the average COD removal efficiencies of SIR1 and SIR2 were 43.1% and 8.4%, respectively (Fig. 2c, d).

a SAA of suspended sludge in SIR1. b SAA of biofilm in SIR1. c SAA of suspended sludge in SIR2. d SAA of biofilm in SIR2. Error bars represent the standard deviation around the mean.

From day 49 to 53, the NH₄⁺-N concentration in the influent of SIR1 increased to 1313.7–1604.9 mg/L (Fig. 1a). The excessive increase in NH₄⁺-N concentration caused the decrease in TIN removal efficiency of SIR1 from 70.3% to 37.3% due to the high concentration of FA (72-88 mg/L), resulting in an increase in the influent NH₄⁺-N concentration of SIR2 from 338.6 mg/L to 897.1 mg/L. Nevertheless, the TIN removal efficiency of the whole system has not been greatly affected, and the TIN removal efficiency can still be maintained at 86.5–93.7%. This was because, although the TIN removal efficiency of SIR1 decreased, TIN can still be removed by the following SIR2 and A²O² units, and the TIN removal efficiency of SIR2 can be maintained at 25.2–41.3%. The influent was ceased on the 54th day and lasted for 3 days to alleviate the inhibitory effect of the high concentration of FA on AnAOB. The SNAD-IFAS process was restored by reducing the influent ammonia nitrogen concentration. From day 57 to 65, the performance of the whole system recovered instantly, with the influent ammonia increasing from 678.7 to 1000.0 mg/L, and the TIN removal efficiency was restored to 94.5%. The SNAD-IFAS hybrid system returned to stabilization in just one week. Actually, it was the speedy adjustment of the hybrid process in the period of break-in that showed excellent practicability since the characteristics of actual wastewater were always complicated and shifty.

Throughout phase II, the A²O² process can ensure that the effluent NH₄⁺-N and NO₂⁻-N concentrations of the hybrid system were kept at a low level. However, the TIN and COD removal efficiency of the A²O² process was relatively low. This was because when the wastewater entered the A²O² unit, most of the bioavailable organic matter in wastewater had been consumed, and most of the remaining organic matter was non-bioavailable organic matter, which makes the denitrification activity and COD removal efficiency of the A²O² unit low. The total nitrogen in the effluent is further reduced if a small additional carbon source is supplemented. In summary, when the undiluted THP-AD liquor was treated by the SNAD-IFAS hybrid process with the average NLR of 0.85 kg N/(m³·d), the average NRR of 0.76 kg N/(m³·d) was maintained, which was higher than the PN/A process^12,13. The average removal efficiencies of NH₄⁺-N, TIN, and COD were 94.0%, 89.9%, and 66.8%, respectively. The removal process of TIN and COD in the THP-AD liquor mainly occurred in SIR1 and SIR2.

Specific anammox activity in SNAD-IFAS reactors

From the above results, most of the nitrogen removal in the SNAD-IFAS hybrid process was completed by SIR1 and SIR2. Therefore, it was necessary to detect the activity of AnAOB, which played a leading role in nitrogen removal in the SNAD-IFAS reactors. Through a series of batch tests, the results showed that strong anammox activity was detected on the biofilm of SIR1 and SIR2, and the SAA of SIR1 and SIR2 was 68.72 and 42.15 mg N/(g VSS·h), respectively (Fig. 3b, d). The SAA inside the SNAD-IFAS process was 7-11 times higher than that inside the PN/A process (6.0 mg N/(g VSS·h))²⁷, which ensured that the SNAD-IFAS process had a higher nitrogen removal loading rate than the PN/A process. No significant anammox activity was detected in the suspended sludge of the two SNAD-IFAS reactors (Fig. 3a, c). However, strong aerobic ammonia oxidation activity can be detected in the suspended sludge of SIR1 and SIR2, and the SAOB can reach 28.04 and 3.55 mg NO₂⁻-N/(g VSS·h), respectively (Fig. S1a, c). The biofilm of the two SNAD-IFAS reactors had weaker aerobic ammonia oxidation activity, with the SAOB of SIR1 and SIR2 being 2.36 and 0.52 mg NO₂^–-N/(g VSS·h), respectively (Fig. S1b, d).

The above specific activity results indicated that AnAOB mainly grew on the biofilm and AOB mainly grew on the suspended sludge. The main agents responsible for nitrogen removal in the SNAD-IFAS process are AnAOB and AOB. This result was consistent with previous studies^21,28,29. Besides, the SAA of the SNAD-IFAS reactors was much higher than previous research reports^18,29,30,31. Two main reasons can be used to explain this phenomenon. One was that the NLR of SNAD-IFAS reactors was much higher than previous studies, and the influent NH₄⁺-N concentration can reach 1347 mg/L, which can provide sufficient substrate for AOB and AnAOB. AnAOB usually had high anammox activity under higher influent substrate conditions³². The other was that a large amount of AnAOB was retained in the reactor to prevent the elapse of AnAOB by adding the biological carriers. The biomass concentration was also a key factor leading to the super high anammox activity³².

Shifts in compositions of microbial community and nitrogen metabolism bacteria

High-throughput sequencing analysis was carried out to further investigate the variation of microbial community structure in the SNAD-IFAS reactors. As shown in Table S1, a total of 257,248 effective sequences were obtained from six samples of suspended sludge and biofilm. These effective sequences were clustered into OTUs, and 3,431 OTUs were obtained. The bacteria coverage of six samples described as the Good’s value was between 99.6% and 99.8% (Table S1). The Ace index and Simpson index were used to analyze the community richness and diversity of suspended sludge and biofilm from different reactors. Compared with inoculated sludge (S0, B0), the Ace index decreased, but the Simpson index increased in SIR1 (S1, B1) and SIR2 (S2, B2), indicating that the microbial richness and diversity in SIR1 and SIR2 were reduced (Table S1). This suggested that non-adapted bacteria in inoculated sludge were eliminated because of changes in the water quality characteristics of the influent. These results agreed with previous studies^33,34. Venn diagrams can be used to tally the number of unique and common OTUs in multiple samples, as well as to visually represent the similarity and overlop of the samples’ OTU composition. According to Venn diagrams (Fig. S2), the unique OTU numbers of the inoculated sludge S0 sample and B0 sample were 360 and 623, respectively. The unique OTU numbers of S1, B1, S2, and B2 samples were 71, 22, 65, and 16, respectively. The inoculated sludge samples had the largest number of unique OTUs, indicating the highest bacterial diversity, which was also consistent with the above Simpson index results. In addition, the different contaminant concentrations of influent in different reactors resulted in a different number of unique OTUs. The unique OTUs in SIR1 were higher than those in SIR2, both in suspended sludge and biofilm. The influent types had a noticeable impact on the microbial community and functional characteristics³⁵.

Apart from generating alterations in microbial richness and diversity, the compositions of microbial community also changed. Bacteroidetes, Patescibacteria, and Fimicutes decreased from 24.44%, 24.03%, and 8.53% in S0 to 13.48%, 0.44%, and 0.99% in S1, respectively (Fig. 4a). The same changes could also be observed in biofilms (B0, B1). However, Planctomycetes, Chloroflexi, and Deinococcus-Thermus significantly increased in SIR1 and SIR2 compared to inoculated sludge. Chloroflexi and Deinococcus-Thermus increased from 1.55% and 1.79% in S0 to 29.08% and 23.26% in S1, respectively. Planctomycetes increase from <0.01% and 9.52% in S0 and B0 to 1.88% and 22.35% in S1 and B1, respectively. Proteobacteria (20.51–38.09%) were dominant in inoculated sludge, SIR1 and SIR2 (Fig. 4a). Proteobacteria were widely detected in a variety of anammox-based processes such as PN/A process and SNAD process^14,21,27. The variations in microbial communities were also mediated using the principal component analysis (PCA) according to all the sequences at OTU levels for qualitative analysis of clustering behavior. The microbial community in both suspended sludge and biofilm was changing after treating THP-AD liquor (Fig. 4b). According to the matrix distances, the biggest differences were between seeding sludge (S0, B0) and activated sludge in both SNAD-IFAS reactors (S1, S2, B1, B2). Further, this evidence was analyzed using a ternary diagram at genus level (Fig. 4c, d). Here, genera from the family Burkholderiaceae, Saprospiraceae, and Ruminococcaceae were significantly enriched in S0 and B0, whereas Truepera, Limnobacter, Nitrospira, Turneriella, Denitratisoma, Thermomonas, SM1A02, Candidatus Kuenenia, and Candidatus Brocadia were mostly detected in S1, S2, B1, and B2. Previous study reported that Limnobacter, Nitrospira, Denitratisoma, and SM1A02 were the common companion bacteria of AnAOB in anammox-based processes^21,28. Truepera and Thermomonas were main decomposers for organic matters^36,37. These results indicated that when the SNAD-IFAS process was used to treat the THP-AD liquor, the microbial community of the inoculated sludge changed dramatically. Some microorganisms adapted to the THP-AD liquor could grow in the SNAD-IFAS reactors, while some microorganisms that were not suitable for the environment were eliminated from the SNAD-IFAS reactors, thus forming a specific microbial community structure in the SNAD-IFAS reactors.

a Community structure at phylum level. b Principle coordinate analysis over long-term operation. c, d Ternary plot representing the relative occurrence of individual genus (circles) in S1, B1 and S2, B2 compared with S0, B0 (genera enriched in different compartments are colored by the taxonomy of the families and the size of the circles is proportional to the mean abundance in the community). e Community structure at genus level. S0: inoculated suspended sludge; S1: SIR1 suspended sludge; S2: SIR2 suspended sludge; B0: inoculated biofilm; B1: SIR1 biofilm; B2: SIR2 biofilm.

Studies have shown that many key nitrogen removal microorganisms belong to Proteobacteria and Planctomycetes, such as Nitrosomonas, Thauera, and Ca. Kuenenia, the model bacteria of nitrogen removal. Therefore, the genus-level compositions of the microbial community were discussed in detail (Fig. 4e, Table 2). The relative abundance of AnAOB (Ca. Kuenenia and Ca. Brocadia) was very low in suspended sludge (<0.01%, S1, S2), whereas it was high in biofilms of the SNAD-IFAS reactors (13.49–20.94%, B1, B2). These results were consistent with the SAA experiment in “Reactor configuration and experimental procedure”. Previous studies also demonstrated that AnAOB mainly existed in the biofilm rather than in the suspended sludge during the operation of the SNAD-IFAS and SNADRP-SBBR processes^21,28. The relative abundance of AnAOB in the SNAD-IFAS process was much higher than that in one stage or two stage PN/A process for treating the THP-AD liquor (10–17%)^11,14. Moreover, Ca. Kuenenia and Ca. Brocadia were the dominant AnAOB in the biofilms of inoculated sludge. However, Ca. Brocadia decreased from 2.66% (B0) to 1.65% (B1), while Ca. Kuenenia increased from 6.22% (B0) to 19.29% (B1), indicating that the dominant AnAOB in SIR1 was Ca. Kuenenia. In contrast, in SIR2, Ca. Brocadia increased from 1.65% (B1) to 4.45% (B2), while Ca. Kuenenia decreased from 19.29% (B1) to 9.04% (B2). The composition of the anammox bacterial community can be significantly influenced by the unique niches that various environments offer. There is a considerable correlation between the composition of the anammox bacterial community and certain environmental factors, including pH, temperature, the contents of nitrogen species, organic carbon and salinity^38,39. A previous study also has demonstrated that Ca. Kuenenia preferred to live in high dissolved inorganic nitrogen level conditions⁴⁰. Consequently, due to the high dissolved inorganic nitrogen level in SIR1, Ca. Kuenenia dominated in SIR1 and had a high relative abundance. However, as the inorganic nitrogen concentration decreased in SIR2, the relative abundance of Ca. Kuenenia decreased, whereas the relative abundance of Ca. Brocadia, which is highly adaptable to low concentrations, increased. The dominant AnAOB in SIR2 was Ca. Kuenenia and Ca. Brocadia. But overall, Ca. Kuenenia was still the dominant AnAOB in the SNAD-IFAS process. In addition, Nitrosomons and Nitrospira appeared in SIR1 and SIR2 (Table 2). According to reports, Nitrosomons and Nitrospira are the major AOB and NOB in anammox-based process^{11,14,19,21,27}. Nitrosomons and Nitrospira were mainly present in the suspended sludge of both SIR1 and SIR2. The relative abundance of Nitrosomons in SIR1 was higher than that in SIR2, whereas the relative abundance of Nitrospira in SIR1 was lower than that in SIR2. This result explained the reason why the nitrate in the effluent of SIR2 was higher than that in the effluent of SIR1 (Fig. 1d). Although the relative abundance of Nitrosomons was very low in the SNAD-IFAS process, the activity of AOB was high (Fig. S1) and had played an important role in anammox-based process^11,14,21,27. The relative abundance of Nitrospira was also lower than that in PN/A process²⁷. Simultaneously, bacteria related to heterotrophic denitrification were found to be enriched in the identified genera, including Thauera, Denitratisoma, Thermomonas, Arenimonas, and Dokdonella (Table 2). Thauera is a common genus of denitrifying bacteria in wastewater treatment plants, and some studies have demonstrated that Thauera is the main partial denitrification bacteria^27,41. Denitratisoma is a type of denitrifying bacteria that often detected in SNAD-IFAS process²¹. Furthermore, Thermomonas, Arenimonas, and Dokdonella are also found to be a heterotrophic denitrifying bacterium and have been observed in partial nitrification-denitrification system for treating landfill leachate and simultaneous nitrification and denitrification process for the treatment of domestic wastewater^42,43. The relative abundances of denitrifying bacteria in SIR1 and SIR2 were 24.75% and 29.43%, respectively. The denitrifying bacteria mostly existed on biofilm in both SIR1 and SIR2, which was also consistent with our previous study²¹. The enrichment of these denitrifying bacteria in SNAD-IFAS reactors further improved the TIN removal efficiency, resulting in a higher TIN removal efficiency than the PN/A process for treating THP-AD liquor. These denitrifying bacteria might also have a certain contribution to the removal of COD in the THP-AD liquor.

Co-occurrence network analysis and organic metabolic functions

In the bioreactor, ecosystem stability depends on relationships between co-existing individuals of microbial communities as well as on the composition of the community as a whole⁴⁴. Co-occurrence networks show more complexity in resistance and resilience of microbial populations than community composition⁴⁵. The co-occurrence patterns (at the genus level) revealed that there were series of positive correlations in bacterial communities of SNAD-IFAS reactor (Fig. 5a). This result indicated that when bacterial communities were exposed to THP-AD liquor containing high concentration of ammonium and complex organics, they inclined to cooperate and weaken competition. The denitrification core genera, Candidatus Kuenenia and Candidatus Brocadia, displayed positive correlations with the cluster of nodes represented by some heterotrophic bacteria, including norank Anaerolineaceae, norank AKYH767, norank Desulfarculaceae, norank SBR1031, norank SJA-28, and Limnobacter. Previous studies reported that most species of the Anaerolineae were responsible for the decomposition of carbohydrates and protein in THP-AD liquor, alleviating the inhibition of complex organics on AnAOB^27,46. The enrichment of Limnobacter was able to consume oxygen and organic matter under organic conditions after long-term operation²¹. Therefore, these heterotrophic bacteria can help AnAOB resist adverse environmental factors such as oxygen and organics. In addition, there were more genera associated with Nitrosomonas in the co-occurrence network, and the relationship was more complex (Fig. 5a). Although the relative abundance of Nitrosomonas (0.21–1.67%) was much lower than other bacteria, Nitrosomonas was central genus in the co-occurrence network. This ecological network characteristic of AOB also explained why predecessors emphasized the importance of AOB in the PN/A process during the treatment of THP-AD liquor^14,27. On the one hand, AOB can provide the important substrate (nitrite) for AnAOB, and on the other hand, AOB occupies an important position in the microbial ecological network and maintains close contact with other microorganisms (Fig. 5a).

a Network analysis based on genus correlation analysis with a 50% cutoff of co-occurrence. The number of the nodes represents the number of genera. Various node colors indicate different phylum. The size of the nodes represents the abundance of genera. The color of the edges indicates the positive (red) and negative (green) of the correlation given Speraman. b, c Sankey diagram indicating the relative contribution of predicted enzymes (ECs) by PICRUSt2.

More importantly, to resist and survive various environmental stresses, which represents an evolutionary challenge for microorganisms, complex defensive mechanisms are necessary for microbial communities⁴⁷. In fact, the THP-AD liquor contains complex organic matters, which would affect the activity of AOB and AnAOB²⁷. However, during the operation of the SNAD-IFAS process, the activity of AOB and AnAOB was high, and the average removal efficiency of COD could reach 43.1%, which was higher than the PN/A process (Figs. 2, 4). This result showed that, compared with the PN/A process, a part of the refractory organics might be degraded in the SNAD-IFAS process. The degradation mechanism of refractory organics might be attributed to the co-metabolism process of organics. Therefore, PICRUSt2 analysis was employed to predict potential co-metabolism mechanisms of organics in the SNAD-IFAS process. As shown in Fig. 5b, the predicted pathways associated with degradation of refractory organics in suspended sludge and biofilm mainly include pyridine degradation, dioxin degradation, cyclohexane degradation, urease degradation, and fatty acid (FAME) degradation. An overall high proportion for enzymes (gabD, EC: 1.2.1.16, 1.2.1.79, 1.2.1.20) involved in the pyridine degradation pathways in suspended sludge and biofilm. The succinate-semialdehyde dehydrogenase (gabD) was mostly involved in the pyridine degradation process’s following stages⁴⁸. Moreover, Truepera was significantly enriched in suspended sludge (S1, S2) and biofilm (B1, B2). Especially in suspended sludge, the relative abundance of Truepera in inoculated suspended sludge was only 1.78%, rapidly increased to 23.36% (S1) and 21.48% (S2), and became the dominant bacteria in suspended sludge (Fig. 4e, Fig. S3). Previous studies reported that Truepera was the dominant species in the biodegradation system treating coking wastewater containing pyridine and an electricity-assisted bio-photodegradation system for high-concentration pyridine removal^49,50. Thus, in this study, the relatively high abundance of Truepera in the SNAD-IFAS system contributed to the removal of pyridine in THP-AD liquor. Apart from the enrichment of specific strains, co-metabolic processes were also effective pathways for pyridine degradation. Some of the carbon-hydride degrading enzymes, including beta-glucosidase (EC: 3.2.1.21) and beta-galactosidase (EC: 3.2.1.23), were anticipated to be present in the SNAD-IFAS process (Fig. 5c). Both enzymes can create easily broken-down small molecules like glucose. Glucose can supply a lot of carbon nutrients and improve energy provision, which encourages the production of extracellular enzymes and/or their secretion, increasing the variety of the microbial community for the co-metabolic of pyridine^51,52. Meanwhile, the creation and secretion of non-specific enzymes that can degrade both substrates can be stimulated by structural analogs of the targeted refractory compounds⁵³. For instance, ammonia can trigger non-specific enzymes, like ammonia monooxygenase, that contribute to the oxidation of refractory materials⁵⁴. Thus, the high ammonia concentration within the SNAD-IFAS system might also promote pyridine co-metabolism. In addition, enzymes that biodegrade with polychlorinated dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs), polychlorinated dibenzofurans (PCB) were also predicted in large numbers, as were enzymes that degrade with cyclohexane, uric acid, fatty acids, and proteins (Fig. 5b)^{55,56,57,58,59}, explaining the high COD removal efficiency of the system. Importantly, these organic degradation enzymes can eliminate the influence of complex organic matter in the system on AOB and AnAOB, and the presence of urease can also decompose urea to provide ammonia nitrogen matrix for AOB and AnAOB. In summary, the complex microbial ecological network formed by the bacteria and the co-metabolism process of refractory organic matter ensure the stable operation of the process.