Full-scale Applications of a Sulfur-Based Autotrophic Denitrification Biofilter for Polishing Secondary Effluent: Annual Performance, Improved Winter Performance, and Microbial Community Succession

Jia-Min Xu; Hao-Ran Xu; Chong-Gang Liang; Ze-Run Cheng; Xu-Chen Ba; Gui-Jiao Zhang; Da-Heng Ren; Na Zhang; Jia-Qiang Lv; Yi-Lu Sun; Aijie Wang; Hao-Yi Cheng

doi:10.1016/j.eng.2025.08.003

Engineering ›› 2026, Vol. 62 ›› Issue (7) :303 -315. DOI: 10.1016/j.eng.2025.08.003

Research

research-article

Full-scale Applications of a Sulfur-Based Autotrophic Denitrification Biofilter for Polishing Secondary Effluent: Annual Performance, Improved Winter Performance, and Microbial Community Succession

Author information +

History +

PDF (7455KB)

Abstract

Sulfur-based denitrification (SADeN) technology is a cost-saving and low-carbon alternative for treating organic-deficient wastewater. However, the use of this technology in full-scale applications is still limited. Crucial concerns for biological processes, such as changes in seasonal performance and the ways in which it responds to low temperatures, have yet to be studied in real SADeN processes. Herein, two SADeN biofilters in parallel (downflow, with a design water treatment capacity of 8000 m³·d⁻¹ for each) were systematically investigated over a period of more than 400 days. Seasonal variations in denitrification rates were observed, with an average difference of up to 2 times between summer and winter. The use of thiosulfate was verified as an efficient strategy to improve the performance of SADeN biofilters in winter, which was found to have an overstoichiometric enhancement effect (OSEE), with 24.79%-331.50% more nitrate removal than that calculated according to thiosulfate dosages. Analyzing the biofilter vertically revealed that the OSEE occurred because thiosulfate rapidly consumed dissolved oxygen in the upper zone of the bed and activated the electron flux of the sulfur-based reactive filler (SReF) in the lower zone. Microbial community analysis further suggested that this increase in electron flux may be associated with the abundance recovery of sulfur autotrophic denitrifiers and the stability recovery of the microbial eco-networks. This study offers a paradigm for the full-scale application of SADeN technology in real wastewater treatment plants, providing an in-depth understanding of the role of dosing thiosulfate in tackling low-temperature challenges.

Graphical abstract

Keywords

Sulfur autotrophic denitrification / Full-scale application / Thiosulfate / Overstoichiometric enhancement effect / Microbial community seasonal succession

Highlight

• Annual performance of full-scale SADeN biofilters was reported for the first time.

• Dosing thiosulfate against low-temperature stress enhanced 3-fold denitrification.

• Thiosulfate rapidly removed DO and amplified the electron flux of S⁰-based filler.

• Seasonal variations drove microbial community succession in SADeN biofilters.

• Thiosulfate improved autotrophic denitrifiers and widened their econiche in winter.

Cite this article

Download citation ▾

Jia-Min Xu, Hao-Ran Xu, Chong-Gang Liang, Ze-Run Cheng, Xu-Chen Ba, Gui-Jiao Zhang, Da-Heng Ren, Na Zhang, Jia-Qiang Lv, Yi-Lu Sun, Aijie Wang, Hao-Yi Cheng. Full-scale Applications of a Sulfur-Based Autotrophic Denitrification Biofilter for Polishing Secondary Effluent: Annual Performance, Improved Winter Performance, and Microbial Community Succession. Engineering, 2026, 62 (7) : 303-315 DOI:10.1016/j.eng.2025.08.003

登录浏览全文

4963

注册一个新账户忘记密码

1. Introduction

Global nitrogen fluxes currently exceed planetary thresholds, causing severe aqueous eutrophication and ecological risks [1]. To mitigate nitrogen discharge into aquatic systems, stringent wastewater discharge regulations have been enforced in various regions, including Europe (Directive 91/271/EEC) [2], the United States (Clean Water Act), and China (GB2002-18918). Previous studies have indicated that nitrate levels in untreated municipal wastewater typically fall within the range of 10-50 mg·L⁻¹ NO₃⁻-N [3], [4], [5], underscoring the necessity for effective denitrification strategies in wastewater treatment plants to mitigate nitrogen pollution before discharge. However, for large-scale wastewater treatment plants (WWTPs) that process thousands of tonnes of wastewater daily, simply modifying the existing activated sludge (A/O) process often falls short of meeting new stringent standards due to its limited improvement potential. Achieving low effluent total nitrogen (TN) levels typically necessitates significantly increasing the recycle ratio, leading to high operational costs. The maximum recycle ratio that can be achieved in full-scale plants is 4:1, making it challenging for most WWTPs to reduce effluent TN to less than 5 mg·L⁻¹ [6]. As discharge policies become more stringent and urban populations continue to grow, municipal WWTPs face substantial challenges in expanding their treatment capacity while managing spatial constraints and reducing energy consumption.

To address these challenges, sulfur-based autotrophic denitrification (SADeN) has emerged as a promising alternative to its heterotrophic counterpart, demonstrating significant potential for treating nitrate-bearing secondary effluent in a cost-effective manner, with low sludge yield and a lower carbon footprint [7], [8], [9]. Despite its potential, most SADeN studies have been limited to lab- or pilot-scale setups [5,10], whereas full-scale studies remain scarce. In previously reported full-scale SADeN studies, studies were carried out over a relatively short period (less than 120 days) and at warm water temperatures (above 20 °C) [11,12]. Given that SADeN is a biological process, its performance is inevitably affected by seasonal temperature variations [13]. In particular, low winter temperatures may significantly deteriorate denitrification efficiency, posing challenges for WWTPs in meeting discharge standards. Nevertheless, to the best of our knowledge, changes in seasonal performance and methods of responding to low temperatures have yet to be studied in real SADeN processes at the full scale.

For counteracting the adverse effects of low temperatures on SADeN performance, the effectiveness of dosing soluble sulfur compounds, such as sulfide [14], thiocyanate [15], and thiosulfate [16,17], has been validated in laboratory-scale studies. Thiosulfate is nontoxic and does not have limitations in use, transportation, or storage [18,19]; therefore, it is an ideal candidate in real SADeN projects. In our previous study, we not only demonstrated that the denitrification performance of SADeN could be significantly improved by the addition of thiosulfate at low temperatures but also that the increased nitrate removal was greater than that predicted according to the dosage of thiosulfate [20]. This phenomenon, named herein the overstochiometric enhancement effect (OSEE), would be very interesting to consider in a real SADeN project, as it may help to reduce the use of thiosulfate. To this end, the OSEE has not yet been systematically investigated.

Based on the above concerns, in this study, two parallel-operating SADeN biofilters (with a designed treatment capacity of 8000 m³·d⁻¹ for each) were systematically investigated for a period of over 400 days, covering the entire year of 2021 and the winter of early 2022. The aims of this study were to ① explore the seasonal performance variations of the SADeN biofilter; ② verify the effectiveness of dosing thiosulfate in mitigating the deterioration of winter performance and elucidate the underlying mechanisms of OSEE; and ③ characterize the microbial community succession driven by temperature changes and thiosulfate dosing. The findings from this study will not only provide practical insights for optimizing the operation of full-scale SADeN biofilters but also deepen our understanding of the roles of thiosulfate in tackling low-temperature challenges.

2. Materials and methods

2.1. Biofilter construction

This wastewater treatment experiment was conducted at an industry zone WWTP in Zhejiang Province, China, employing two sets of long-term stabilized operating SADeN biofilter units (referred to as SADeN-BF^A and SADeN-BF^B in this study) with identical specifications and configurations. A photograph, schematic diagram, and inflow arrangement of the SADeN biofilter are depicted in Fig. 1. The size of each SADeN biofilter tank was 7.0 m in length × 7.0 m in width × 5.7 m in height. In the upper part of the biofilter tank, two 0.4-m wide water diversion grooves with evenly spaced filter heads formed a 0.5-m high distribution zone to ensure uniform inflow. The working volume of the SADeN biofilter was designed to be 112.7 m³ by filling with a sulfur-based reactive filler (SReF). The SReF (diameter 3-6 mm), as described in our patent [21], was synthesized mainly from S⁰ and natural siderite and was collaboratively produced by Beijing Zhongchi Bize Environmental Technology Co., Ltd., China (trademark is known as ThiocreF^®). For the vertical construction of the SADeN biofilter, the vertical direction consisted of the supporting layer (filter brick, 0.2 m; 4-10 mm diameter cobblestones, 0.3 m; and 1-3 mm diameter quartz sand, 0.2 m), packed bed layer (4-10 mm diameter SReF, 2.3 m), and prefiltration water zone (2.2 m), from bottom to top.

2.2. Experimental parameters for biofilters

Throughout the experiment, which was conducted from January 1, 2021 to March 1, 2022, the biofilter received secondary biochemical effluent as the influent in a downflow mode (Fig. 1). Further details regarding the characteristics of the influent can be found in Table S1 in Appendix A. The biofilter operations were divided into two phases (Table 1). In phase I (days 1-378), the long-term denitrification performance of two sets of SADeN-BFs operating under identical operational parameters was investigated to evaluate the impact of influent NO₃⁻-N loading fluctuations (0.31-0.95 kilogram nitrogen per cubic meter per day (kg-N·m⁻³·d⁻¹)) and temperature variations (13-30 °C) on denitrification performance. Throughout the operation, the inlet flow rate was regulated by the influent valve, with an empty-bed contact time (EBCT) of 0.29-0.67 h. In phase II, industrial sodium thiosulfate (Na₂S₂O₃·5H₂O, purity > 98%) was dosing into the influent of the SADeN-BF^A to understand its influence on denitrification performance under low-temperature conditions (13.5-18.2 °C). Specifically, the thiosulfate (S₂O₃²⁻) dosages were set at 17.0, 0, 8.5, 17.0, 25.5, and 0 mg·L⁻¹ (the dosing quality was adjusted on the basis of the actual daily influent water volume), corresponding to the subphases within phase II as follows: II-1 (days 379-390), II-2 (days 391-401), II-3 (days 402-408), II-4 (days 409-414), II-5 (days 415-420), and II-6 (days 421-425). The selection of the thiosulfate dosage was based on a combination of theoretical calculations (Eq. (3)) and empirical optimization, which were designed to meet the most stringent nitrogen discharge policies in China: TN < 5 mg·L⁻¹ (DB5301/T43-2020). Moreover, SADeN-BF^B was operated continuously during this phase without the addition of thiosulfate, serving as a control set. During the operation of the SADeN biofilters, backwashing was initiated when the water level surpassed the influent weir. The routine backwashing procedure consisted of a 3-min air wash for nitrogen removal, followed by a 5-min combined air-and-water wash, and finally a 5-min water wash to eliminate aged biofilms and prevent clogging [5].

2.3. Chemical analysis

To obtain samples from the extended biofilter beds at different heights, a custom steel vertical sampling rod (schematic diagram in Fig. 1(c)) was used. It consists of a long, embedded rod and a matching outer sleeve. The embedded rod features a hollow section (inner diameter, 10 cm; height, 8 cm) only at the sampling sites (50, 100, 150, and 200 cm bed height), whereas the rest remains solid, and the samples can be collected by rotating the outer sleeve. The water quality of the influent and effluent was represented by samples from 0 and 230 cm bed heights, respectively. All the water samples were filtered through 0.45 μm membrane filters before analysis. Anions (NO₃⁻-N, NO₂⁻-N, S₂O₃²⁻, and SO₄²⁻), alkalinity, and chemical oxygen demand (COD) were measured using ultraviolet spectrophotometry (DR6000, HACH, USA) with HACH preinstalled reagents (HACH). Dissolved oxygen (DO) and pH were measured via a multiparameter water quality analyzer (Multi 3630, WTW GmbH, Germany) equipped with a DO sensor (DO 925-3, WTW GmbH) and a pH sensor (pH-Electrode SenTix 940-3, WTW GmbH).

2.4. Microbial analysis

Biofilm-attached SReF samples were collected from both sets of SADeN-BFs via the custom-made steel vertical sampling rod during phase I (A_190d^50/100/200 and B_190d^50/100/200) in summer and during phase II-1 (A_378d^50/100/200, A_390d^50/100/200*, and B_390d^50/100/200), II-2 (A_395d^50/100/200), II-5 (A_420d^50/100/200* and B_420d^50/100/200), and II-6 (A_425d^50/100/200) in winter. Here, samples labeled A and B were taken from SADeN-BF^A and SADeN-BF^B, respectively; the subscript number indicates the sampling time; the superscript 50/100/200 indicates the distance of the sampling site from the influent interface at the top of the biofilter packing layer to the bottom, cm; and “*” indicates the winter sample with thiosulfate dosing. Three parallel samples were collected from each sampling site. A total of 81 biofilm samples formed on the SReF were subsequently collected for DNA extraction and sequencing. Detailed information regarding the biological preprocedures can be found in our prior reports [22]. The total DNA of each biofilm sample was extracted using the E.Z.N.A.^® Soil DNA Kit (Omega Biotek, USA). The V3-V4 hypervariable regions of the bacterial 16S ribosomal RNA (rRNA) gene were amplified with 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) primers using the GeneAmp 9700 thermocycler (ABI, USA). The sequence data were clustered at a 97% similarity cutoff, and taxonomic assignments were performed using the Majorbio Cloud online platform. Microbial eco-network topology analysis was conducted using the integrated network analysis pipeline [23]. Details of the data processing and analysis can be found in Appendix A.

2.5. Calculation methods

2.5.1. Denitrification rate

The NO₃⁻-N removal rate $ \left(R_{\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}\right)$ was calculated as follows (Eq. (1)):

(1)$ R_{\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}=\frac{\left(C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}-C_{\mathrm{Ef}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}\right) \times Q}{V_{\mathrm{W}}} \times \frac{24}{1000}$

where $ C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}$ represents the influent nitrate concentration (mg·L⁻¹), $ C_{\mathrm{Ef} \mathrm{-NO}_{3}{ }^{-}-\mathrm{N}}$ represents the effluent nitrate concentration, (mg·L⁻¹), Q represents the influent flow (L·h⁻¹), and V_w represents the volume of the SReF within the biofilter bed (L).

2.5.2. Estimation of the OSEE

Thiosulfate was dosed into the SADeN biofilter during the cold winter, where it fulfilled two key roles: A fraction was consumed for DO removal (Eq. (2)) [24], whereas the remaining fraction facilitated nitrate removal (Eq. (3)) [17]. The OSEE (%) induced by thiosulfate addition was estimated using Eqs. (4), (5), (6).

(2)

{S_{2} O}_{3}^{2 -} + 1.65 O_{2} + 1.12 H_{2} O + {0.333CO}_{2} + {0.022HCO}_{3}^{-} + {0.071 {NH}_{4}}^{+} \to 0.071 C_{5} H_{7} O_{2} N + 2.05 H^{+} + {2 {SO}_{4}}^{2 -}

(3)

{S_{2} O}_{3}^{2 -} + {1.24 {NO}_{3}}^{-} + 0.11 H_{2} O + {0.45 {HCO}_{3}}^{-} + {0.09 {NH}_{4}}^{+} \to 0.09 C_{5} H_{7} O_{2} N + 0.40 H^{+} + 0.62 N_{2} + {2 {SO}_{4}}^{2 -}

(4)

C_{Δ DO}^{Ac} = (C_{In-DO}^{Dt} - C_{Ef-DO}^{Dt}) - (C_{In-DO}^{Un} - C_{Ef-DO}^{Un})

(5)$ C_{\Delta \mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Ac}}=\left(C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Dt}}-C_{{\mathrm{Ef}-\mathrm{NO}_{3}}^{-}-\mathrm{N}}^{\mathrm{Dt}}\right)-\left(C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Un}}-C_{{\mathrm{Ef}-\mathrm{NO}_{3}}^{-}-\mathrm{N}}^{\mathrm{Un}}\right)$

(6)$ \mathrm{OSEE}=\left(\frac{C_{\Delta \mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Ac}} \times \frac{112}{1.24 \times 14}}{C_{\mathrm{In}-\mathrm{S}_{2} \mathrm{O}_{3}{ }^{2-}}-\mathrm{C}_{\Delta \mathrm{DO}}^{\mathrm{Ac}} \times \frac{112}{1.65 \times 32}}-1\right) \times 100 \%$

where $ C_{\Delta \mathrm{DO}}^{\mathrm{Ac}}$ represents the actual difference in DO removal between thiosulfate-dosed and undosed conditions, whereas $ C_{\Delta \mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Ac}}$ represents the actual difference in NO₃⁻-N removal (mg·L⁻¹). $ C_{\mathrm{In}-\mathrm{DO}}^{\mathrm{Dt}}$ and $ C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Dt}}$ denote the influent DO and NO₃⁻-N concentrations under thiosulfate dosing, respectively, whereas $ C_{\mathrm{Ef}-\mathrm{DO}}^{\mathrm{Dt}} $ and $ C_{\mathrm{Ef}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Dt}}$ represent the corresponding effluent concentrations (mg·L⁻¹). Similarly, $C_{\mathrm{In}-\mathrm{DO}}^{\mathrm{Un}} $ and $C_{\mathrm{In}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Un}}$ are the influent concentrations without thiosulfate, respectively, and $C_{\mathrm{Ef}-\mathrm{DO}}^{\mathrm{Un}}$ and $C_{\mathrm{Ef}-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}^{\mathrm{Un}}$ are the corresponding effluent concentrations (mg·L⁻¹). $C_{\mathrm{In}-\mathrm{S}_{2} \mathrm{O}_{3}{ }^{2-}}$ indicates the concentration of thiosulfate dosed into the influent (mg·L⁻¹).

2.5.3. Electron donor and acceptor equilibrium

The vertical electron transfer rates for DO, TN, thiosulfate (S₂O₃²⁻), and elemental sulfur (S⁰) were calculated using the following equations:

For DO as the electron acceptor:

(7)

e_{acceptor-DO}^{-} = \frac{(C_{H2-DO} - C_{H1 -DO}) \times 4}{32 \times {EBCT}_{H1 – H2}}

For TN (including nitrate and nitrite) as the electron acceptor:

(8)

e_{acceptor-TN}^{-} = \frac{(C_{H2-N {O_{3}}^{-} -N} - C_{H1 -N {O_{3}}^{-} -N}) \times 2}{14 \times {EBCT}_{H1 – H2}} + \frac{(C_{H2-N {O_{2}}^{-} -N} - C_{H1 -N {O_{2}}^{-} -N}) \times 3}{14 \times {EBCT}_{H1 – H2}}

For S₂O₃²⁻ as the electron donor:

(9)

e_{{donor-S}_{2} {O_{3}}^{2 -}}^{-} = \frac{(C_{H2- S_{2} {O_{3}}^{2 -}} - C_{H1 - S_{2} {O_{3}}^{2 -}}) \times 8}{112 \times {EBCT}_{H1 – H2}}

For S⁰ as the electron donor:

(10)

e_{donor- S^{0}}^{-} = e_{acceptor-DO}^{-} + e_{acceptor-TN}^{-} - e_{donor- S_{2} {O_{3}}^{2 -}}^{-}

where e⁻₍_x₎ represents the electron transfer rate related to each substance x per volume of wastewater at a given packed-bed height between H1 and H2 (mole e⁻ per cubic meter per hour (mol-e⁻·m⁻³·h⁻¹)). The parameters H1 and H2 denote specific bed layer heights (e.g., 0, 50, 100, 150, 200, and 230 cm), with H2 being the greater height (cm). The terms $C_{\mathrm{H} 2-\mathrm{DO}}$, $\quad C_{\mathrm{H} 2-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}$, $C_{\mathrm{H} 2-\mathrm{NO}_{2}^{-}-\mathrm{N}}$, and $\quad C_{\mathrm{H} 2-\mathrm{S}_{2} \mathrm{O}_{3}^{2-}}$ represent the concentrations of DO, NO₃⁻-N, NO₂⁻-N, and S₂O₃²⁻, respectively, at height H2, whereas $C_{\mathrm{H} 1-\mathrm{DO}}$, $\quad C_{\mathrm{H} 1-\mathrm{NO}_{3}{ }^{-}-\mathrm{N}}$, $C_{\mathrm{H} 1-\mathrm{NO}_{2}^{-}-\mathrm{N}}$, and $\quad C_{\mathrm{H} 1-\mathrm{S}_{2} \mathrm{O}_{3}^{2-}}$ correspond to their respective concentrations at height H1 (mg·L⁻¹). EBCT_H1–H2 represents the empty bed contact time (h) for wastewater, as it flows through the bed layer between heights H1 and H2. Here, the total EBCT for the entire bed (0–230 cm) was determined to be 0.3 h, which was based on the average value obtained from actual measurements. According to these equations, the additional electron flux (AEF) of the SReF activated by thiosulfate dosing can be estimated as the percentage increase in $\mathrm{e}_{\text {donor }-\mathrm{S}^{0}}^{-}$ when thiosulfate is dosed compared with the condition without thiosulfate (%).

2.6. Statistical analyses

Statistical analyses were conducted using one-way ANOVA with Tukey’s post-hoc test for multiple comparisons. When data did not meet normality assumptions, the Kruskal-Wallis test was applied. A significance level of p < 0.05 was considered statistically significant.

3. Results and discussion

3.1. Annual denitrification performance of the two full-scale biofilters

Throughout the year, two full-scale sulfur-based autotrophic denitrification biofilters (SADeN-BF^A and SADeN-BF^B) were operational within a WWTP in Zhejiang Province, China (Fig. 1), which is known for its subtropical monsoon climate featuring four distinct seasons. As depicted in Fig. 2(a), the influent NO₃⁻-N loadings (inNLs) for the SADeN-BFs fluctuated year-round, ranging from 0.31 to 0.95 kg kg-N·m⁻³·d⁻¹, reflecting the variability of wastewater in real WWTPs. The annual variation curves (phase I, days 1-378) of effluent NO₃⁻-N concentrations in both biofilters exhibited significant overlap, aligning with a trend nearly opposite to that of water temperature (WT) changes. Specifically, effluent NO₃⁻-N concentrations gradually increased from January as the WT decreased, peaked at the end of February, and then continuously declined to approximately 1 mg·L⁻¹ until the middle of May. This low effluent NO₃⁻-N concentration then remained stable for approximately 180 days (average WT, (25.93 ± 2.10) °C). In October, the effluent NO₃⁻-N concentration again started to increase with decreasing WT. Overall, NO₃⁻-N removal varied seasonally, with average efficiencies in spring, summer, autumn, and winter recorded as (69.32% ± 26.06%), (96.84% ± 3.84%), (80.18% ± 15.90%), and (43.13% ± 11.59%) for SADeN-BF^A, and (67.21% ± 24.76%), (93.23% ± 4.90%), (77.62% ± 12.65%), and (46.31% ± 12.24%) for SADeN-BF^B, respectively.

To gain further insight into the effects of the WT on denitrification performance, the annual NO₃⁻-N removal rates of the two biofilters were calculated and categorized by collecting data from phase I. As shown in Fig. 2(b-i), the NO₃⁻-N removal rates of the SADeN-BF^A and SADeN-BF^B samples were not significantly different from each other at any point (p > 0.05), with average values of (0.20 ± 0.05), (0.30 ± 0.09), (0.49 ± 0.09), and (0.53 ± 0.11) kg-N·m⁻³·d⁻¹ for those of the WT samples ranging from 13-16, 16-20, 20-25, and 25-30 °C, respectively. Similarly, the adverse effects of low WTs on denitrification rates have also been documented in some laboratory-scale SADeN reactors [20]. Unlike these laboratory-scale studies, which have been conducted for shorter durations, our long-term investigation revealed greater variability in denitrification rates at higher WTs, particularly at 25-30 °C, suggesting the influence of additional factors beyond water temperature in engineering practice. This observation aligns with the constructed linear correlations between denitrification rates and inNLs, indicating that inNLs, fluctuations also play a crucial role in denitrification performance (Fig. 2(b-ii)).

3.2. Promotion of the denitrification rate by the addition of thiosulfate and the OSEE

In phase II (days 379-425), both SADeN biofilters were continuously operated during winter, with thiosulfate dosing exclusively into the influent of SADeN-BF^A to investigate its impact on denitrification. Following the initial dosing of 17.0 mg·L⁻¹ thiosulfate in phase II-1 (Fig. 2(a)), effluent NO₃⁻-N concentrations were noticeably lower in SADeN-BF^A than in SADeN-BF^B ((3.89 ± 0.52) vs (7.13 ± 0.73) mg·L⁻¹). Upon cessation of thiosulfate dosing in phase II-2, effluent NO₃⁻-N concentrations in SADeN-BF^A remained lower for a certain interval (approximately 10 days) until gradually becoming comparable to those in SADeN-BF^B. In phase II-3, redosing thiosulfate, even at a reduced dosage of 8.5 mg·L⁻¹, resulted in a noticeable decrease in effluent NO₃⁻-N concentrations again, stabilizing below 4 mg·L⁻¹ within 5 days. By increasing the thiosulfate dose stepwise from 8.5 to 25.5 mg·L⁻¹ (phases II-3 to II-5), SADeN-BF^A achieved significantly higher nitrate removal efficiencies than did SADeN-BF^B ((58.78% ± 8.78%) to (63.05% ± 5.77%) vs (27.66% ± 7.43%)). Similarly, the nitrate removal rate calculated from SADeN-BF^A increased from (0.38 ± 0.04) to (0.55 ± 0.07) kg-N·m⁻³·d⁻¹ (Fig. 2(b-iii)), resulting in a 1.90- to 2.80-fold increase compared with that of SADeN-BF^B with no thiosulfate dosing.

When thiosulfate is introduced into the SADeN biofilter, it serves two primary functions: consuming DO (Eq. (2)) and facilitating nitrate removal (Eq. (3)). To assess its denitrification enhancement, a mass balance analysis was conducted in phase II. Given that aerobic respiration generally outcompetes denitrification for electrons [25], thiosulfate was assumed to be first utilized for DO removal to establish a material balance. Since no additional electron donors were introduced, the remaining thiosulfate was attributed to denitrification. Moreover, the contribution of S⁰ to DO and nitrate removal was considered constant, as observed in the absence of thiosulfate (Fig. 3(a), thiosulfate = 0 mg·L⁻¹). Notably, regardless of the thiosulfate dosage (8.5-25.5 mg·L⁻¹), the theoretical nitrate removal (i.e., 0.66-3.11 mg·L⁻¹ NO₃⁻-N) remained consistently lower than the actual nitrate removal achieved (2.85-3.88 mg·L⁻¹ NO₃⁻-N). This disparity indicated an OSEE (24.79%-331.50% more nitrate removal) induced by thiosulfate dosing, implying that thiosulfate not only directly participated in denitrification but also likely stimulated additional biochemical pathways or enhanced existing processes beyond its expected stoichiometric contribution.

3.3. Vertical profiles of biofilter performance in response to seasonal changes and thiosulfate dosing

To reveal the response of the SADeN biofilter to seasonal changes and thiosulfate dosing, vertical profiles of the biofilter were analyzed throughout the winter and summer. Here, a distinct seasonal difference was observed, with notably higher NO₃⁻-N and DO concentrations in winter than in summer, accompanied by increased NO₂⁻-N accumulation (Fig. 3(b)). This difference in the vertical profile diminished with increasing thiosulfate dose. During winter, thiosulfate not only reduces NO₃⁻-N but also alleviates NO₂⁻-N accumulation, facilitating more complete denitrification [16,17]. As depicted in Fig. 3(b), the applied thiosulfate dosages ensured that no residues remained in the effluent, with depletion observed at 8.5, 17.0, and 25.5 mg·L⁻¹ within the 50-100, 100-150, and 150-200 cm bed layers, respectively. Notably, thiosulfate consumption was most rapid within the 0-50 cm bed layer, which coincided with a relatively high DO removal efficiency (Winter-S₂O₃²⁻: 49.53%-58.19% vs Winter, 28.03%), whereas NO₃⁻-N removal was concentrated in the middle-to-lower bed layers. Thiosulfate preferentially consumes DO in the upper bed layers of biofilters because of the higher energy yield of aerobic respiration than of denitrification, making DO the preferred electron acceptor in these layers [25]. Additionally, DO competes with nitrate by reversibly inhibiting denitrification enzyme activity and modulating related gene expression [26,27], further favoring oxygen as the terminal electron acceptor.

Furthermore, vertical variations in electron flux were analyzed to elucidate how thiosulfate dosing enhanced denitrification alongside the observed OSEE. The total electron flux from thiosulfate at different bed heights could be derived from its vertical concentration profile (Fig. 3(b)). We quantified the electron flux absorbed by acceptors (NO₃⁻-N and DO) on the basis of electron transfer equilibrium principles and inferred the electron flux supplied by the solid SReF (primarily S⁰) by subtracting the electron contribution of thiosulfate. The overall vertical distribution revealed a lower electron transfer rate in the Winter bed than in the summer bed (Figs. 4(a) and (b)), with the denitrification hot zone concentrated in the 150–200 cm bed. As shown in Figs. 4(c)–(e), thiosulfate dosing enabled rapid DO elimination in the upper bed layers (0–50 cm) by supplementing additional electron donors while shifting the denitrification hot zone upward (100–150 cm) and expanding its range (0–230 cm). More importantly, the electron contribution from SReF increased to varying degrees after thiosulfate dosing, as evidenced by the increase in $\mathrm{e}_{\text {donor }-\mathrm{S}^{0}}^{-}$ to the right of the yellow dashed line. This may explain the OSEE induced by thiosulfate, as its introduction activated the AEF from the SReF (indicated by the red lines), resulting in an overstoichiometric increase in denitrification rates beyond what thiosulfate alone could contribute. Notably, under oxygen-limited conditions, partial oxidation of thiosulfate to S⁰ and sulfate may have occurred [28]. This could lead to an underestimation of the electron contribution from thiosulfate to denitrification, as S⁰ could further participate in denitrification. Consequently, the AEF effect of SReF may be overestimated. In the 150–200 cm bed, where thiosulfate was depleted, the observed electron flux increase for denitrification could only originate from SReF, confirming the AEF activation effect. Specifically, in this zone, the AEF increase ratios for SReF at thiosulfate dosages of 8.5, 17.0, and 25.5 mg·L⁻¹ were 94.13%, 81.08%, and 49.41%, respectively.

3.4. Dynamic succession features of the microbial community

Over different operational times (days 190, 378, 390, 395, 420, and 425), 81 biofilm samples were collected from two SADeN biofilters at three bed heights (50, 100, and 200 cm) for high-throughput sequencing, with the aim of revealing the microbial community structure and succession. First, multidimensional principal coordinate analysis (multi-PCoA) was used to categorize all the samples into three groups (Fig. 5(a)), explaining 82.552% of the total bacterial community variation. The clear intergroup dispersion and intragroup clustering indicated that temperature was a key driver of community differentiation. Notably, the Winter-S₂O₃²⁻ samples clustered closer to the summer samples, suggesting that thiosulfate dosing induced a summer-like microbial composition in winter.

To further analyze the functional species distribution, the relative occurrence of the top ten genera was examined using a ternary plot (Fig. 5(b)). Thiobacillus, Ferritrophicum, and Sulfuricurvum, the key sulfur autotrophic denitrifying bacteria (SADBs) [29], [30], [31], were more prevalent in Summer (27.44%) but were suppressed in Winter samples (8.17%). However, Thiobacillus and Ferritrophicum were markedly more abundant in Winter-S₂O₃²⁻ than in Winter samples (17.42% and 2.71% vs 5.42% and 1.72%, p < 0.05; Table S2 in Appendix A), highlighting the facilitative role of thiosulfate in SADB proliferation under cold stress (21.73% vs 8.17%). Conversely, three genera (Parachlamydia, Sphingopyxis, and unclassified_o_Saccharimonadales; circled in blue) were significantly more abundant in Winter and Winter-S₂O₃²⁻ than in Summer samples (p < 0.001; Table S2), likely due to their inherently psychrophilic traits [32], [33], [34].

Temporal and spatial succession patterns were further analyzed. As shown in Fig. 5(c), Thiobacillus (21.36%-23.66% in summer, as a representative) and Dokdonella (12.70%-16.57%) dominated distinct biofilters. While Thiobacillus utilizes various reducible sulfur species (including S⁰, S²⁻, and S₂O₃²⁻) for denitrification [30,35], Dokdonella is an obligate aerobe with incomplete denitrification capabilities, reducing NO₃⁻-N into NO₂⁻-N [36,37]. Both genera declined in winter due to low-temperature inhibition [38]. However, Thiobacillus rebounded significantly upon thiosulfate dosing and redosing in winter (18.71% and 16.13% in A_390d* and A_420d*, respectively) but decreased after thiosulfate withdrawal (A_390d* to A_395d and A_420d* to A_425d), underscoring its strong dependence on thiosulfate under low-temperature stress. Conversely, Dokdonella remained unaffected, likely due to its inability to utilize thiosulfate or S⁰ for growth [39,40].

Spatially, Thiobacillus abundance consistently increased from 50 to 200 cm bed depth during both summer (18.33% to 25.25%) and winter (4.36% to 6.10%), whereas Dokdonella abundance showed the opposite trend (17.53% to 13.14% vs 7.41% to 5.67%). This pattern likely resulted from declining DO levels along the bed height (Fig. 3(b)), favoring facultative anaerobic Thiobacillus over strictly aerobic Dokdonella. Notably, after thiosulfate dosing, Thiobacillus increased significantly in the upper biofilter layers, which coincided with substantial DO reduction. This correlation was likely attributed to certain species of Thiobacillus utilizing thiosulfate while consuming DO [41]. Although thiosulfate can react with O₂ abiotically, this process is extremely slow because of unfavorable molecular orbital interactions [42]. Experimental evidence further supports this, as studies have shown that even under continuous aeration in sterile conditions, thiosulfate oxidation by molecular oxygen is minimal [43]. Thus, microbial-mediated oxidation is the dominant pathway for thiosulfate turnover. Overall, thiosulfate dosing increased the relative abundance of genera exceeding 1% across bed heights, even at the 200 cm bed where thiosulfate was depleted. This finding aligned with the observed increase in OSEE (Fig. 3(a)), suggesting that thiosulfate facilitated the electron flux of SReF, thereby supporting autotrophic microbial growth.

3.5. Associations between environmental factors and functional bacteria

Redundancy analysis (RDA)/canonical correspondence analysis (CCA) analysis was conducted to elucidate the key environmental factors shaping the distribution and abundance of keystone microorganisms, with a particular focus on temperature (T), influent DO (InDO), removed DO (ReDO), thiosulfate dosing (ThioD), and nitrate removal rates (NaRR). As shown in Fig. 5(d), vectors representing Thiobacillus, Ferritrophicum, and Sulfuricurvum formed very small acute angles with the NaRR vector (nearly in the same direction), reflecting a strong positive correlation between these SADBs and the denitrification rate. Additionally, the acute angles and proximity of these bacterial vectors to the T vector underscore the importance of relatively warm temperatures in promoting SADB proliferation. Notably, all SADBs were negatively correlated with the identified psychrotrophs (including Parachlamydia, Sphingopyxis, and Saccharimonadales), which exhibited heterotrophic growth features (details available in Table S2) and were significantly more abundant in winter than in summer (9.49% vs 1.34%; Fig. 5(b)). This finding revealed that, in addition to low-temperature stress, SADBs in winter may also face increased competition from heterotrophic psychrotrophs, which tend to thrive by consuming SADBs under nutrient-scarce conditions. Furthermore, these psychrotrophs exhibited a positive correlation with InDO, which aligns with previous studies that highlighted the role of elevated DO concentrations in promoting their growth [44]. The negative correlation between these psychrotrophs and the ReDO vector suggested that lower DO concentrations may limit their growth. Vertical distribution analysis further supported the hypothesis that psychrotrophs were less abundant in the Winter-S₂O₃²⁻ bed than in the Winter bed (e.g., 200 cm bed, 6.21% vs 9.09%; Fig. 5(e)). Moreover, the observed vertical dwcline in psychrotroph abundance aligns with DO variations (e.g., Winter-S₂O₃²⁻ 50 to 200 cm bed, 8.99% to 6.21%), likely reinforcing the importance of thiosulfate-driven DO reduction in alleviating the pressure from psychrotrophs to SADBs.

In response to thiosulfate dosing, all SADBs presented increased abundances and notable shifts, as evidenced by the strong positive correlations, close alignment, and acute angles between their vectors and the ThioD vector (Fig. 5(d)). The introduction of thiosulfate, a soluble electron donor with superior bioavailability compared with that of solid S⁰ [17], is undeniably a pivotal factor in enhancing the proliferation of SADBs. Furthermore, the strong positive correlation among the ThioD, ReDO, and NaRR vectors indicated that thiosulfate not only facilitated DO removal but also potentially limited the growth of heterotrophic psychrotrophs, thereby allowing SADBs to thrive via the autotrophic denitrifying process. The classified statistical results of the microbial nutritional and functional patterns (Fig. 5(e)) demonstrated that following thiosulfate dosing, the abundance of psychrotrophs decreased significantly, especially in the lower bed layers (e.g., 200 cm, 6.21% vs 9.09%; p < 0.05). Conversely, SADBs displayed a markedly different vertical distribution pattern, achieving more than 2.36-fold directional enrichment at the 200 cm bed layer under thiosulfate dosing compared to conditions without dosing (21.75% vs 9.20%; p < 0.05). Importantly, at this depth, thiosulfate was completely consumed (Fig. 3(b)), indicating that solid-phase SReF was instrumental in supporting the growth of these SADBs. SADBs can share sulfur utilization pathways, although the mechanisms by which they leverage solid monomeric sulfur for growth remain inadequately understood. The current literature suggests that SADBs may solubilize solid S⁰ into bioavailable forms, potentially through dissolved zero-valent S⁰ or polysulfides [45], thus acting as dissolved electron donors. Consequently, the observed enrichment of SADBs in the lower bed layers likely elucidates the amplification of electron flux from SReF following thiosulfate dosing (Figs. 4(c)-(e)). This highlights the integral role of thiosulfate in optimizing the microbial structure and enhancing the denitrification efficiency within the SADeN biofilter.

3.6. Microbial eco-network shift in the SADeN biofilter

Seasonal and thiosulfate-driven differentiation in the microbial network was analyzed using the Sparse Correlations for Compositional data (SparCC) method to construct phylogenetic molecular eco-networks (pMENs) (Fig. 6). The key parameters representing the topological structure of the network are summarized in Table S3 in Appendix A. Compared with those in Summer, the Winter eco-network (Fig. 6(b)) presented greater size (total OTUs and edges), connectivity (average clustering coefficient), and complexity (modularity) (Fig. 6(a)), indicating closer intertaxa relationships under low temperature. Similar seasonal profiles have been observed in marine benthic archaeal communities, which are associated with decreased microbial activity and limited bioavailable substrates under colder conditions, necessitating tighter microbial interactions [46]. In contrast, the Winter-S₂O₃²⁻ eco-network (Fig. 6(c)) displayed increased interaction strength (closeness centrality, connectedness, and efficiency), reduced path distances, and greater modularity (0.3245 vs 0.2514) than did the Winter eco-network. These results suggested that thiosulfate dosing alleviated some low-temperature-induced network instability [47], enhancing microbial interactions and stabilizing the community [48].

Modularity, which reflects spatial segregation and resource partitioning in microbial communities [49], was employed to investigate the distribution and connections of SADBs within specific modules. Seasonal changes in the ecological niches of SADBs were observed: SADBs (top ten genera, identified OTUs in Fig. 6) were dispersed across five modules (S-M1, S-M2, S-M4, S-M9, and S-M11) in summer, whereas they aggregated into a single module (W-M1) in winter, indicating intensified resource competition under low-temperature stress. As mentioned earlier, SADBs can utilize solid S⁰ by dissolving it into bioavailable forms (such as dissolved zero-valent S⁰ or polysulfides), depending on biomediated S⁰ reduction or disproportionation processes within local microenvironments [45,50]. In winter, these potential processes may be substantially attenuated due to suppressed microbial activity under colder conditions [38], leading to reduced solubilization of solid S⁰. This insufficiency of bioavailable sulfur likely intensified competition among SADBs, possibly resulting in the displacement of certain SADBs during ecological niche competition (e.g., OTU2399, OTU9941, OTU3602, OTU2907, OTU2264, and OTU2163, which were present in summer but absent in winter). Notably, thiosulfate dosing mitigated this competition by replenishing the sulfur source, as evidenced by the redistribution of SADBs in the Winter-S₂O₃²⁻ eco-network (W-T-M1, W-T-M2, W-T-M3, and W-T-M4). The observed increase in SADBs following thiosulfate dosing is likely associated with enhanced electron flux (Fig. 4(c)-(e)), as these bacteria play a crucial role in eco-network stability and directly consume reductive sulfur species (i.e., electron donors).

3.7. Implications and outlook

This study provides, for the first time, a comprehensive evaluation of the effects of seasonal dynamics on annual SADeN-BF performance in real-world operational settings. The results clearly showed that the effluent NO₃⁻-N concentration fluctuated with variations in water temperature and influent NO₃⁻-N loading. Notably, in the case of low influent NO₃⁻-N loading in summer and high influent NO₃⁻-N loading in winter, the difference in NO₃⁻-N removal can be several times greater (e.g., data recorded from day 210 to 240 vs those recorded from day 360 to 370, as shown in Fig. 2(a)). Notably, for a denitrification biofilter, the effluent NO₃⁻-N concentration is typically expected to be slightly lower than the regulatory limit, which allows the process to cost-effectively meet discharge standards. For this reason, the results underscore the necessity of developing adaptive regulation methods for SADeN-BF that enable precise modulation in the efficiency of denitrification to stabilize effluent NO₃⁻-N concentrations.

In this study, we verified that the use of thiosulfate as an external electron donor is an effective approach to increase the denitrification efficiency of SADeN-BF. Moreover, the presence of an OSEE in the context of thiosulfate dosing was exciting. Since thiosulfate-driven denitrification is more expensive than that driven by elemental sulfur [51], the discovery of OSEE suggests that the actual cost of thiosulfate dosing may be lower than the cost predicted on the basis of stoichiometry. In addition to thiosulfate, other soluble sulfur compounds (such as sulfide and sulfite) may also be effective as external electron donors [51,52]. Future studies should explore potential differences in OSEE when these alternative electron donors are used. In addition to upregulation, downregulation of denitrification efficiency is also essential for stabilizing the effluent NO₃⁻-N concentration. Recently, we proposed a strategy involving adjustment of the bed-immersion ratio, in which the denitrification efficiency can be decreased by simply lowering the bed-immersion ratio [53]. In the future, it will be important to integrate the strategies of external electron donor dosing and bed-immersion ratio variation to enable bidirectional regulation of SADeN-BF denitrification efficiency. In this context, further efforts should be made to develop a control system incorporating models and sensors to guide electron donor dosing and bed-immersion ratio regulation.

4. Conclusion

This study provides a comprehensive full-scale assessment of the SADeN process (8000 m³·d⁻¹) over 400 days, evaluating its denitrification performance and low-temperature resilience. The following key conclusions were drawn:

• The annual application efficacy of the SADeN process was systematically tracked in a real WWTP for the first time. The results confirmed that advanced denitrification could be achieved, with effluent nitrate concentrations exhibiting clear seasonal variations in response to temperature changes, averaging (2.90 ± 2.60) mg·L⁻¹ throughout the year.

• Dosing thiosulfate proved to be an effective strategy for improving denitrification during colder months, resulting in a greater than two-fold increase in nitrate removal. The observed OSEE demonstrated that thiosulfate dosing enhanced the denitrification efficiency beyond the theoretical expectations, offering a critical solution for low-temperature challenges.

• Thiosulfate dosing improved performance via two key mechanisms: ① consuming DO in the upper bed, which helped to limit heterotrophic psychrotrophs and alleviate their competitive pressure on SADBs; and ② serving as a highly bioavailable electron donor, enriching SADBs and amplifying the electron flux from the SCreF, ultimately increasing the denitrification efficiency.

Our findings offer valuable insights into improving secondary wastewater treatment performance, particularly under challenging low-temperature conditions, positioning SADeN technology as a competitive solution for advanced nitrate removal in WWTPs.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Rockström J, Steffen W, Noone K, Persson Å, Chapin III FS, Lambin EF, et al. A safe operating space for humanity. Nature 2009; 461(7263):472-5.

[2]	European Parliament. Proposal for a directive of the European Parliament and of the council concerning urban wastewater treatment (recast). Report. Brussels: European Commission; 2022.

[3]	Rivett MO, Buss SR, Morgan P, Smith JWN, Bemment CD. Nitrate attenuation in groundwater: a review of biogeochemical controlling processes. Water Res 2008; 42(16):4215-32.

[4]	Sahinkaya E, Kilic A, Duygulu B. Pilot and full scale applications of sulfur-based autotrophic denitrification process for nitrate removal from activated sludge process effluent. Water Res 2014; 60:210-7.

[5]	Wang SS, Cheng HY, Zhang H, Su SG, Sun YL, Wang HC, et al. Sulfur autotrophic denitrification filter and heterotrophic denitrification filter: comparison on denitrification performance, hydrodynamic characteristics and operating cost. Environ Res 2021; 197:111029.

[6]	Gao X, Zhang L, Liu J, Zhang Y, Peng Y. First application of the novel anaerobic/ aerobic/anoxic (AOA) process for advanced nutrient removal in a wastewater treatment plant. Water Res 2024; 252:121234.

[7]	Di Capua F, Pirozzi F, Lens PNL, Esposito G. Electron donors for autotrophic denitrification. Chem Eng J 2019; 362:922-37.

[8]	Xu JM, Dong H, Xu HR, Sun YL, Yu Y, Zhang LY, et al. Water flush boosts performance of elemental sulfur-based denitrification packed-bed systems: optimization and mechanisms. Bioresour Technol 2024; 408:131158.

[9]	Xu JM, Sun YL, Yao XD, Zhang GJ, Zhang N, Wang HC, et al. Highly efficient coremoval of nitrate and phosphate driven by a sulfur-siderite composite reactive filler toward secondary effluent polishing. Environ Sci Technol 2023; 57(43):16522-31.

[10]	Sierra-Alvarez R, Beristain-Cardoso R, Salazar M, Gómez J, Razo-Flores E, Field JA. Chemolithotrophic denitrification with elemental sulfur for groundwater treatment. Water Res 2007; 41(6):1253-62.

[11]	Xu Z, Chen X, Li H, Wan D, Wan J. Combined heterotrophic and autotrophic system for advanced denitrification of municipal secondary effluent in full- scale plant and bacterial community analysis. Sci Total Environ 2020; 717:136981.

[12]	Zhou Y, Chen F, Chen N, Peng T, Dong S, Feng C. Denitrification performance and mechanism of biofilter constructed with sulfur autotrophic denitrification composite filler in engineering application. Bioresour Technol 2021; 340:125699.

[13]	Zhang N, Sun YL, Yao BM, Zhang B, Cheng HY. Insight into the shaping of microbial communities in element sulfur-based denitrification at different temperatures. Environ Res 2022; 215:114348.

[14]	Qi X, Han J, Kou Z, Liang P. Supplementary sulfide during inoculation for improved sulfur autotrophic denitrification performance and adaptation to low temperature. Sci Total Environ 2023; 900:166365.

[15]	Broman E, Jawad A, Wu X, Christel S, Ni G, Lopez-Fernandez M, et al. Low temperature, autotrophic microbial denitrification using thiosulfate or thiocyanate as electron donor. Biodegradation 2017; 28(4):287-301.

[16]	Bao HX, Li ZR, Song ZB, Wang AJ, Zhang XN, Qian ZM, et al. Mitigating nitrite accumulation during S0-based autotrophic denitrification: balancing nitrate- nitrite reduction rate with thiosulfate as external electron donor. Environ Res 2022; 204:112016.

[17]	Yang Y, Perez Calleja P, Liu Y, Nerenberg R, Chai H. Assessing intermediate formation and electron competition during thiosulfate-driven denitrification: an experimental and modeling study. Environ Sci Technol 2022; 56 (16):11760-70.

[18]	Di Capua F, Milone I, Lakaniemi AM, Lens PNL, Esposito G. High-rate autotrophic denitrification in a fluidized-bed reactor at psychrophilic temperatures. Chem Eng J 2017; 313:591-8.

[19]	Zou G, Papirio S, Lakaniemi AM, Ahoranta SH, Puhakka JA. High rate autotrophic denitrification in fluidized-bed biofilm reactors. Chem Eng J 2016; 284:1287-94.

[20]	Zhang XN, Zhu L, Li ZR, Sun YL, Qian ZM, Li SY, et al. Thiosulfate as external electron donor accelerating denitrification at low temperature condition in S0-based autotrophic denitrification biofilter. Environ Res 2022; 210:113009.

[21]	Cheng HY, Wang AJ, Peng YZ, Wang SS, Zhu TT inventors; Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, assignee. Nitrogen and phosphorus removal active biological carrier, its preparation method and application. Chinese patent ZL201811113486.4. 2020 Apr 14. [Chinese].

[22]	Xu JM, Zi HY, Xu HR, Zhang YF, Ren DH, Zeng R, et al. Improved efficiency and stability using a novel elemental sulfur-based moving-bed denitrification process. Water Res 2024; 254:121391.

[23]	Feng K, Peng X, Zhang Z, Gu S, He Q, Shen W, et al. iNAP: an integrated network analysis pipeline for microbiome studies. iMeta 2022; 1(2).

[24]	Mora M, López LR, Lafuente J, Pérez J, Kleerebezem R, van Loosdrecht MCM, et al. Respirometric characterization of aerobic sulfide, thiosulfate and elemental sulfur oxidation by S-oxidizing biomass. Water Res 2016; 89:282-92.

[25]	Thauer RK, Jungermann K, Decker K. Energy conservation in chemotrophic anaerobic bacteria. Bacteriol Rev 1977; 41(1):100-80.

[26]	Thomas KL, Lloyd D, Boddy L. Effects of oxygen, pH and nitrate concentration on denitrification by pseudomonas species. FEMS Microbiol Lett 1994; 118(1-2):181-6.

[27]	Gómez M. Effect of dissolved oxygen concentration on nitrate removal from groundwater using a denitrifying submerged filter. J Hazard Mater 2002; 90 (3):267-78.

[28]	Houghton JL, Foustoukos DI, Flynn TM, Vetriani C, Bradley AS, Fike DA. Thiosulfate oxidation by Thiomicrospira thermophila: metabolic flexibility in response to ambient geochemistry. Environ Microbiol 2016; 18(9):3057-72.

[29]	Kodama Y, Watanabe K. Sulfuricurvum kujiense gen. nov., sp. nov., a facultatively anaerobic, chemolithoautotrophic, sulfur-oxidizing bacterium isolated from an underground crude-oil storage cavity. Int J Syst Evol Microbiol 2004; 54(6):2297-300.

[30]	Ravichandra P, Mugeraya G, Rao AG, Ramakrishna M, Jetty A. Isolation of Thiobacillus sp. from aerobic sludge of distillery and dairy effluent treatment plants and its sulfide oxidation activity at different concentrations. J Environ Biol 2007; 28(4):819-23.

[31]	Wan D, Li Q, Liu Y, Xiao S, Wang H. Simultaneous reduction of perchlorate and nitrate in a combined heterotrophic-sulfur-autotrophic system: secondary pollution control, pH balance and microbial community analysis. Water Res 2019; 165:115004.

[32]	Greub G, La Scola B, Raoult D. Parachlamydia acanthamoeba is endosymbiotic or lytic for acanthamoeba polyphaga depending on the incubation temperature. Ann NY Acad Sci 2003; 990(1):628-34.

[33]	Pan Y, Sun RZ, Wang Y, Chen GL, Fu YY, Yu HQ. Carbon source shaped microbial ecology, metabolism and performance in denitrification systems. Water Res 2023; 243:120330.

[34]	Ting L, Williams TJ, Cowley MJ, Lauro FM, Guilhaus M, Raftery MJ, et al. Cold adaptation in the marine bacterium, Sphingopyxis alaskensis, assessed using quantitative proteomics. Environ Microbiol 2010; 12(10):2658-76.

[35]	Yu L, Yuan Y, Chen S, Zhuang L, Zhou S. Direct uptake of electrode electrons for autotrophic denitrification by Thiobacillus denitrificans. Electrochem Commun 2015; 60:126-30.

[36]	Li Y, Zhang J, Chen Q, Cai S, He J, Zhou S, et al. Dokdonella kunshanensis sp. nov., isolated from activated sludge, and emended description of the genus Dokdonella. Int J Syst Evol Microbiol 2013; 63(Pt 4):1519-23.

[37]	Ten LN, Jung HM, Im WT, Oh HW, Yang DC, Yoo SA, et al. Dokdonella ginsengisoli sp. nov., isolated from soil from a ginseng field, and emended description of the genus Dokdonella. Int J Syst Evol Microbiol 2009; 59(8):1947-52.

[38]	Kelly DP, Wood AP. Confirmation of Thiobacillus denitrificans as a species of the genus Thiobacillus, in the beta-subclass of the Proteobacteria, with strain NCIMB 9548 as the type strain. Int J Syst Evol Microbiol 2000; 50 (2):547-50.

[39]	Valle A, Fernández M, Ramírez M, Rovira R, Gabriel D, Cantero D. A comparative study of eubacterial communities by PCR-DGGE fingerprints in anoxic and aerobic biotrickling filters used for biogas desulfurization. Bioprocess Biosyst Eng 2018; 41(8):1165-75.

[40]	Ziegler S, Waidner B, Itoh T, Schumann P, Spring S, Gescher J. Metallibacterium scheffleri gen. nov., sp. nov., an alkalinizing gammaproteobacterium isolated from an acidic biofilm. Int J Syst Evol Microbiol 2013; 63(Pt 4):1499-504.

[41]	Justin B, Kelly DP. Metabolic changes in Thiobacillus denitripcans accompanying the transition from aerobic to anaerobic growth in continuous chemostat culture. J Gen Microbiol 1978; 107(1):131-7.

[42]	Xu Y, Schoonen MAA. The stability of thiosulfate in the presence of pyrite in low-temperature aqueous solutions. Geochim Cosmochim Acta 1995; 59 (22):4605-22.

[43]	Rolia E, Chakrabarti CL. Kinetics of decomposition of tetrathionate, trithionate, and thiosulfate in alkaline media. Environ Sci Technol 1982; 16(12):852-7.

[44]	Zhu L, Xu H, Xiao W, Lu J, Lu D, Chen X, et al. Ecotoxicological effects of sulfonamide on and its removal by the submerged plant Vallisneria natans (Lour.) Hara. Water Res 2020; 170:115354.

[45]	Boyd ES, Druschel GK. Involvement of intermediate sulfur species in biological reduction of elemental sulfur under acidic, hydrothermal conditions. Appl Environ Microbiol 2013; 79(6):2061-8.

[46]	Liu J, Zhu S, Liu X, Yao P, Ge T, Zhang XH. Spatiotemporal dynamics of the archaeal community in coastal sediments: assembly process and co- occurrence relationship. ISME J 2020; 14(6):1463-78.

[47]	Olesen JM, Bascompte J, Dupont YL, Jordano P. The modularity of pollination networks. Proc Natl Acad Sci 2007; 104(50):19891-6.

[48]	Coyte KZ, Schluter J, Foster KR. and stability. The ecology of the microbiome: networks, competition, Science 2015; 350(6261):663-6.

[49]	Li C, Jin L, Zhang C, Li S, Zhou T, Hua Z, et al. Destabilized microbial networks with distinct performances of abundant and rare biospheres in maintaining networks under increasing salinity stress. iMeta 2023; 2(1).

[50]	Franz B, Lichtenberg H, Hormes J, Modrow H, Dahl C, Prange A, et al. Utilization of solid ‘elemental’ sulfur by the phototrophic purple sulfur bacterium Allochromatium vinosum: a sulfur K-edge X-ray absorption spectroscopy study. Microbiology 2007; 153(4):1268-74.

[51]	Wang T, Li X, Wang H, Xue G, Zhou M, Ran X, et al. Sulfur autotrophic denitrification as an efficient nitrogen removals method for wastewater treatment towards lower organic requirement: a review. Water Res 2023; 245:120569.

[52]	Bao HX, Wang HL, Wang ST, Sun YL, Zhang XN, Cheng HY, et al. Response of sulfur-metabolizing biofilm to external sulfide in element sulfur-based denitrification packed-bed reactor. Environ Res 2023; 231:116061.

[53]	Zhang G, Xu HR, Liao W, Xu JM, Zhang N, Ren D, et al. Bed-immersion-ratio variation as an efficient strategy to regulate denitrification efficiency directionally in elemental sulfur packed-bed reactors. Water Res 2025; 272:122941.