Effect of Carbon Source on Endogenous Partial Denitrification Process: Characteristics of Intracellular Carbon Transformation and Nitrite Accumulation

Abstract

This study focused on the start-up and operating characteristics of the endogenous partial denitrification (EPD) process with different carbon sources. Two sequencing batch reactors (SBRs) with sodium acetate (SBR1^#) and glucose (SBR2^#) as carbon sources were operated under anaerobic/oxic (A/O) and anaerobic/anoxic/oxic (A/A/O) modes successively for 240 d. The results showed that COD removal efficiency reached 85% and effluent COD concentrations were below 35 mg/L in both SBRs. The difference was that faster absorption and transformation of sodium acetate was achieved compared to glucose (COD removal rate (CRR) was 7.54 > 2.22 mgCOD/(L·min) in SBR1^# compared to SBR2^#). EPD could be started up with sodium acetate and glucose as carbon sources, respectively, and desirable high nitrite accumulations were both obtained at influent NO₃⁻−N (NO₃⁻-N_inf) increased from 20 to 40 mg/L with nitrate-to-nitrite transformation ratio (NTR) and specific NO₃⁻-N deduction rate (r_Na) of 88.4~90% and 2.41~2.38 mgN/(gVSS·h), respectively. However, at NO₃⁻-N of 50~60 mg/L, both the NTR and r_Na in SBR1^# were higher compared to SBR2^# (86.5% > 83.9% and 1.58 > 1.20 mgN/(gVSS·h), respectively). Hereafter, when NO₃⁻-N was increased by 70~90 mg/L, lower NTR and r_Na were observed in SBR1^# than in SBR2^# (72% and 78%, 1.16 and 1.32 mgN/(gVSS·h), respectively). Additionally, similar internal carbon transformations were observed to drive EPD for NO₂⁻−N accumulation, especially for higher and faster carbon transformation with sodium acetate as carbon source compared to glucose. However, precise control of anoxic time as the peak point of nitrite (T_Ni,max) was still the key to achieve high NO₂⁻−N accumulation.

1. Introduction

The novel autotrophic biological nitrogen removal process of anaerobic ammonium oxidation (anammox) has broad application prospects in sustainable wastewater treatment, due to its advantages in saving aeration energy, low sludge yield, high-efficiency nitrogen removal and zero organic carbon source consumption, etc. [1,2,3]. During the anammox process, ammonia (NH₄⁺−N) and nitrite (NO₂⁻−N) could be directly converted to nitrogen gas (N₂) under anaerobic conditions using Equation (1) [4]. However, the stable acquisition of nitrite still restricted the present popularization and application of anammox.

NH₄⁺ + 1.32NO₂⁻ + 0.066HCO₃⁻ + 0.13H⁺ → 1.02N₂ + 0.26NO₃⁻ + 0.066CH₂O_2.5N_0.15 + 2.03H₂O

(1)

Endogenous partial denitrification (EPD) overcame the bottleneck of simultaneous achievements of high NO₂⁻−N accumulation and deep high organic carbons removal during the denitrification process, providing a broader application prospect for anammox [5]. During the EPD process, glycogen-accumulating organisms (GAOs) could fully absorb the organic carbons in sewage through the glycolysis pathway (EMP) under anaerobic conditions and efficiently store it as poly-β-hydroxy-alkanoates (PHAs); hereafter, the stored PHAs could be drive denitrification, cell growth and glycogen recovery under anoxic conditions [6]. Hence, a stable and high nitrite accumulating rate could be achieved by enhancing the denitrification performance of GAOs with reasonable controlling of anaerobic and anoxic reaction times [7].

However, GAOs have always been regarded as the “scavenger” in enhanced biological phosphorus removal (EBPR) systems, since they are competing for a carbon source with phosphorus-accumulating organisms (PAOs) and thus leading to the deterioration of dephosphatation [8]. Recently, many studies focused on the physiological and metabolic mechanisms of GAOs [9,10], its gene sequence [11], morphological characteristics and factors influencing its activity [12], aiming to inhibit its growth activity in EBPR systems. Few studies turned to improving the partial denitrification performance of GAOs and their coupling characteristics with anammox bacteria [1]. Specially, carbon source had important influences on the synthesis and composition of intracellular PHAs, and then played a decisive role in the carbon and nitrogen metabolic pathways [13]. It is of great significance to investigate the variations in GAOs activity with different carbon sources and ascertain its effect on the partial denitrification performance.

For now, acetic acid has been widely used as external carbon source in municipal wastewater treatment plants [14], while glucose was the most studied (except for other volatile fatty acids (VFAs)) but reached inconsistent comments on its influence on different functional nutrient removal processes, especially for anaerobic phosphorus release and oxic phosphorus uptake [15]. Zhang, Guisasola and Baeza [16] confirmed that using glucose as carbon source in an EBPR system would lead to deterioration of phosphorus release and uptake due to the rapid proliferation of GAOs, and Liu, Cheng, Ren, Zhang, Wang, Wan and Lv [17] found that glucose as carbon source could achieve stable phosphorus removal. It is representative to study the influence of these two carbon sources of acetic acid and glucose on the EPD process of GAOs.

In this study, the effect of carbon source on EPD was investigated in two sequencing batch-activated sludge reactors (SBRs) feeding sodium acetate and glucose, respectively. The start-up and long-term operation stability of these two EPD-SBR systems were studied to identify the different characteristics of carbon source transformation, intracellular carbon composition, nitrate removal and nitrite accumulation. It is hoped to provide theoretical basis and test data support for the performance optimization of the EPD process treating nitrate-containing wastewater.

2. Materials and Methods

2.1. Reactors and Operation Procedures

Two laboratory-scale SBRs (working volume: 12 L) were employed in the current study (Figure 1). One SBR (named SBR1^#) was fed with sodium acetate as carbon source, while the other (named SBR2^#) was fed with glucose. Both of the SBRs were equipped with pH/DO meters, air diffusers and mechanical stirrers and operated at 20 ± 1 °C by automatic temperature control heating systems. The dissolved oxygen (DO) was controlled at 1.5–2.0 mg/L by adjusting a gas flowmeter. The pH was kept at 6.8–7.0 by adding diluted HCl solutions (0.1 mmol/L).

Both SBR1^# and SBR2^# were operated for 240 d with 12 different operating phases according to the operating mode and influent substrate concentration (

Table 1. Operating conditions and influent substrates at different phases of SBR1^# and SBR2^#.

Phase	Operating Time (d)		Operating Mode		Duration (min)						Influent (mg/L)
					Anaerobic		Anoxic		Oxic		COD		NO3−-N		PO43−-P
	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#	SBR1#	SBR2#
1	1~30		A/O		180		—		150		150		—		0
2	31~60		A/O		180		—		150		200		—		0
3	61~90		A/O		180		—		150		250		—		0
4	91~120		A/O		180		—		150		300		—		0
I	121~132		A/A/O		180		30	30	50		150	200	20		4
II	133~146		A/A/O		180		40	45	50		200	200	30		4
III	147~160		A/A/O		180		60	60	50		250	250	40		4
IV	161~174		A/A/O		180		80	90	50		250	250	50		4
V	175~188		A/A/O		180		90	120	50		250	300	60		4
VI	189~212	189~202	A/A/O		180		100	140	0	30	300	350	70		4
VII	213~226	203~220	A/A/O		180		120	150	0	20	350	350	80		4
VIII	227~240	221~240	A/A		180		150	160	—		350	350	90		4

). In phase 1–4 (1–120 d), the two reactors were both operated under anaerobic/oxic condition and fed with limited phosphorus concentration to enrich GAOs but inhibit PAOs growth, whereas in phase I-VII (121–226 (SBR1^#)/220 (SBR2^#) d), the operation mode was adjusted to anaerobic/anoxic/oxic to start up EPD and hereafter changed to anaerobic/anoxic to improve EPD performance in phase VIII (227 (SBR1^#)/221 (SBR2^#)-240 d) without limited-phosphorus operation. Specially, in each operation cycle (6 h), 3 L of wastewater was introduced to each reactor in the first 5 min of anaerobic phase resulting in a drainage ratio of 37.5%; 150–200 mL of mixed liquor was discharged before settling to maintain the mixed liquor suspended solid (MLSS) of about 4.0 g/L.

2.2. Wastewaters and Seed Sludge

Synthetic wastewater containing a carbon source (sodium acetate or glucose) and NH₄⁺−N (NH₄Cl) was composed of tap water with a specific chemical oxygen demand (COD) concentration of 150–350 mg/L and NH₄⁺−N of about 40 mg/L (Table 1). The detailed composition was (/L): 0–0.024 g of KH₂PO₄, 0.08 g of MgCl₂∙6H₂O, 0.045 g of MgSO₄∙7H₂O, 0.144 g of KCl, 0.02 g of CaCl₂∙2H₂O and 1.2 mL of trace element solution [18]. The trace element solution contained (/L): 1.5 g of FeCl₃∙6H2O, 0.03 g of CuSO₄∙5H₂O, 0.12 g of MnCl₂∙4H₂O, 0.06 g of Na₂MoO₄∙2H₂O, 0.12 g of ZnSO₄∙7H₂O, 0.15 g of CoCl₂∙6H₂O, 0.18 g of KI, 0.15 g of H₃BO₃ and 10 g of EDTA. Synthetic nitrate wastewater fed to the reactor contained NO₃⁻−N (NaNO₃) with a concentration of 20–90 mg/L (Table 1).

2.3. Methods for Chemical Analysis

All samples were taken every day in a fixed operation cycle of 9:00 a.m. to 11:00 a.m. and then filtered through 0.45 µm filter paper before analysis. NH₄⁺−N, NO₂⁻−N, NO₃⁻−N, PO₄³⁻-P, TN, MLSS and MLVSS were analyzed according to standard methods [19]. COD was analyzed using a COD quick-analysis apparatus (Lian-Hua Tech. Co., Ltd., 5B-3A, Beijing, China), and pH value and DO concentration were measured by a pH/DO meter (pH/oxi340i, WTW Company, Munich, Germany). Freeze-dried biomass was used to measure polyhydroxyalkanoates (PHAs) and glycogen (Gly). PHAs were determined by the sum of poly-β-hydroxybutyrate (PHB) and poly-β-hydroxyvalerate (PHV), which were analyzed as previously reported [20]. Glycogen was analyzed as previously reported [9].

2.4. Calculation of Nitrate-to-Nitrite Transformation Ratio NTR

Nitrate-to-nitrite transformation ratio (NTR) is defined as the proportion of nitrite production in nitrate removal during the endogenous denitrifying process at the anoxic phase in SBR1^# or SBR2^# (Equation (2)).

$$\mathrm{NTR}(\%)=(\mathrm{N}{\mathrm{O}}_{2,\mathrm{T}}^{-}-\mathrm{N}{\mathrm{O}}_{2,\mathrm{i}}^{-})/(\mathrm{N}{\mathrm{O}}_{3,\mathrm{i}}^{-}-\mathrm{N}{\mathrm{O}}_{3,\mathrm{T}}^{-})\times 100\%$$

(2)

where NO₂⁻_,i and NO₃⁻_,i are the NO₂⁻−N and NO₃⁻−N concentrations at the beginning of the anoxic phase, mg/L, respectively; while NO₂⁻_,T and NO₃⁻_,T are their concentrations at anoxic time of T_Ni,max; T_Ni,max is the time when NO₂⁻−N concentration reaches the maximum at the anoxic phase, min.

3. Results and Discussion

3.1. Effects of Carbon Source on Anaerobic COD Removal in SBR1^# and SBR2^# under A/O Operation Mode

The variations in COD concentration and removal efficiency of SBR1^# and SBR2^# (with sodium acetate and glucose as carbon source, respectively) under A/O operation mode were analyzed to illustrate the anaerobic COD removal performance during the EPD process, as shown in Figure 2.

In Phase 1 (1~30 d), the influent COD (COD_inf) concentration was 150 mg/L, the COD concentrations at the end of the anaerobic phase (COD_an,e) of SBR1^# and SBR2^# were 78.3 and 88.0 mg/L, respectively, with effluent concentrations of 37.7 and 35.4 mg/L. the anaerobic COD removal efficiency values were only 47.7% and 41.3%, respectively, indicating a poor COD removal performance. Additionally, the effluent COD (COD_eff) concentrations were 30.6 and 27.2 mg/L in SBR1^# and SBR2^#, respectively, illustrating that COD removal occurred predominantly at the oxic phase of both SBRs.

In Phase 2 (31~60 d), by increasing COD_inf to 200 mg/L, COD_an,e of SBR1^# decreased from 58.0 mg/L to 29.0 mg/L with anerobic COD removal efficiency increased from 71.0% to 85.5%. No significant change in COD concentration occurred at the oxic phase, indicating that the improved COD removal performance occurred in the anaerobic phase. As for SBR2^#, COD_an,e and anaerobic COD removal efficiency were stably maintained at 42.5 mg/L and 78.7%, indicating that increasement of influent COD has no obvious effect on COD removal.

In Phase 3 (61~90 d), COD_inf was increased to 250 mg/L. Concentrations of COD_an,e and COD_eff were both maintained at 25.0 mg/L, with anaerobic COD removal efficiency reaching 86.1%. In SBR2^#, COD_an,e diminished from 58.0 to 29.0 mg/L, accompanied by anaerobic COD removal efficiency having risen from 77.1% to 89.2%. In this phase, excellent COD removal performance was achieved in both SBR1^# and SBR2^# with sodium acetate and glucose, respectively, served as carbon source.

In Phase 4 (91~120 d), COD_inf was further increased to 300 mg/L. SBR1^# reached a stable anaerobic COD removal performance, with average COD_an,e and COD_eff concentrations of 33.8 and 30.0 mg/L and COD removal efficiency of 87.0%. This confirmed the stably maintained excellent anaerobic COD removal performance of SBR1^# even with increased influent COD (150 to 300 mg/L). However, in SBR2^#, the average COD_an,e was as high as 80.7 mg/L, with anaerobic COD removal efficiency reduced to 67.8%, indicating that high influent concentrations of COD had a greater effect on anaerobic COD removal with glucose served as the carbon source than with sodium acetate.

3.2. Mechanisms of COD Removal in SBR1^# and SBR2^# with Different Carbon Sources

In order to explore the mechanism of COD removal in SBR1^# and SBR2^# with different carbon sources, variations in COD concentration, removal efficiency and removal rate during typical operation cycles on Day 52 were analyzed, as shown in Figure 3.

During the anaerobic phase (0~15 min) of SBR1^#, COD concentration decreased rapidly from 200 to 86.8 mg/L, with COD removal efficiency of 56.6% and COD removal rate (CRR) reaching 7.54 mgCOD/(L·min). During 16~180 min, COD concentration continuously decreased to 32.2 mg/L, with CRR declined from 4.60 to 0.93 mgCOD/(L·min). The COD removal efficiency reached 83.9% finally (Figure 3A). The results indicated that COD was mainly removed in the first 15 min at the anaerobic phase of SBR1^# with sodium acetate served as the carbon source.

Different from SBR1^#, at the anaerobic phase of SBR2^#, COD decreased from 200 to 153.3 mg/L in the first 15 min. COD removal efficiency was only 23.3% with a low CRR of 3.11 mgCOD/(L·min). Whereas, in the later 16~180 min, COD slightly decreased to 32.2 mg/L, with CRR stabilized at 2.22 mgCOD/(L·min) and COD removal efficiency reaching 77.4% (Figure 3B). Thus, it could be seen that when glucose was used as the carbon source, the COD removal performance was inferior to that of sodium acetate, which might be attributed to the obvious differences in internal carbon source conversion between different carbon sources [21].

In the oxic phase, the COD concentrations in SBR1^# and SBR2^# were both stabilized at 31.5 and 34.5 mg/L, respectively, which further proved the main COD removal in the anaerobic phase with sodium acetate and glucose used as carbon sources.

3.3. Effects of Carbon Sources on Endogenous Partial Denitrification Performance in SBR1^# and SBR2^# under A/A/(O) Operation Mode

In order to explore the endogenous partial denitrification performance of SBR1^# and SBR2^# under A/A/O operation mode, using sodium acetate and glucose as sole carbon sources, variations in NO₃⁻−N and NO₂⁻−N concentrations and NTR in the two SBRs were analyzed, as shown in Figure 4.

In Phase I–III (121~160 d), when COD_inf concentrations of SBR1^# and SBR2^# were increased from 150 and 200 mg/L, respectively, to 250 mg/L and anoxic influent NO₃⁻−N concentrations (NO₃⁻−N_inf) were both increased from 20 to 40 mg/L, the denitrification performances of both the two SBRs were improved obviously. Specifically, effluent NO₂⁻−N (NO₂⁻−N_eff) concentrations increased from 12.7 and 18.2 mg/L to 35.6 and 37.2 mg/L, respectively, with effluent NO₃⁻−N (NO₃⁻−N _eff) concentrations both lower than 0.3 mg/L. In addition, the T_Ni,max of SBR1^# were 20, 40 and 60 min during phases 1~3, respectively, and 30, 45 and 60 min of SBR2^#, respectively, which elucidated the delayed duration achieving the maximum accumulation of NO₂⁻−N under higher NO₃⁻−N_inf, NTR, and specific NO₃⁻−N reduction rate (r_NaR) in SBR1^# and SBR2^# were, respectively, 88.4% and 2.41 mgN/(gVSS·h) and 90.0% and 2.38 mgN/(gVSS·h), indicating that high NO₃⁻−N removal and NO₂⁻−N accumulation were both achieved in the two SBRs without being affected by carbon sources.

Similar results were obtained in Phase IV (161~174 d), when NO₃⁻−N_inf were both increased to 50 mg/L in SBR1^# and SBR2^# with COD_inf maintained at 250 mg/L. The average NO₂⁻−N_eff in the two SBRs was 43.3 and 42.2 mg/L, respectively, with both NO₃⁻−N_eff values nearly zero. Meanwhile, T_NO2--N,max was delayed to 80 min and 90 min for SBR1^# and SBR2^#, respectively. The NTR and r_NaR in SBR1^# and SBR2^# were 88.4% and 1.48 mgN·(gVSS·h), 90% and 1.2 mgN/(gVSS·h), respectively. These results indicated that 50 mg/L of NO₃⁻−N_inf and different carbon sources have no influence on the endogenous partial denitrification performance.

In Phase V (175~188 d), NO₃⁻−N_inf was increased to 60 mg/L. NO₃⁻−N _eff, NO₂⁻−N_eff and NTR were 0.2 mg/L, 47.2 mg/L and 82.3%, respectively, in SBR1^# with T_Ni,max decreased from 100 to 85 min and r_NaR increased from 1.34 to 1.67 mgN/(gVSS·h), indicating the enhanced EPD performance. As for SBR2^#, COD_inf was increased to 300 mg/L, NO₃⁻−N _eff, NO₂⁻−N_eff and NTR were 0.02 mg/L, 48.9 mg/L and 81.4%, respectively, with T_Ni,max stabilized at 120 min and r_NaR of 1.2 mgN/(gVSS·h). In this phase, sodium acetate showed superiority in facilitating EPD start-up and NO₂⁻−N accumulation compared to glucose.

NO₃⁻−N_inf was increased to 70 mg/L in Phase VI, with COD_inf increased to 300 mg/L in SBR1^# and 350 mg/L in SBR2^#. In this phase, the denitrification performance of SBR1^# was obviously affected in days 200~206, with NO₂⁻−N_eff rapidly decreased from 51.0 to 28.2 mg/L and NO₃⁻−N_eff increased from 1.4 to 5.4 mg/L, NTR was only 43.7%; hereafter, NO₃⁻−N_eff, NO₂⁻−N_eff and NTR recovered back to 1.65 mg/L, 54.7 mg/L, and 78.2%, respectively, and T_Ni,max reached stability at 100 min, benefited from the operation mode change to A/A in days 207~212. As for SBR2^# (189~202 d), the EPD performance was almost unaffected but stabilized with average NO₂⁻−N_eff and NTR of 56.2 mg/L and 75%. Additionally, r_NaR of SBR1^# and SBR2^# were 1.12 and 1.04 mgN/(gVSS·h), respectively. Thus, better NO₂⁻−N accumulation performance via EPD could be achieved using glucose as the carbon source, compared to sodium acetate, under high NO₃⁻−N_inf of 70 mg/L.

NO₃⁻−N_inf was continuously increased to 80 mg/L in Phase VII, with COD_inf increased to 350 mg/L in SBR1^#. Stable EPD performances were achieved in both SBRs. NO₃⁻−N_eff, NO₂⁻−N_eff and NTR in SBR1^# were 0.02, 53.9 mg/L and 71%, with 0.9, 62.5 mg/L and 79.71% in SBR2^#, respectively, and r_NaR values were, respectively, 1.30 and 1.26 mgN/(gVSS·h). With NO₃⁻−N_inf further increased to 90 mg/L, the EPD performances of SBR1^# and SBR2^# were both significantly affected in Phase VIII, showing as NO₃⁻−N_eff increased to 7.1 and 9.9 mg/L, respectively, NO₂⁻−N_eff decreased to 48.6 and 50.89 mg/L, and r_NaR and NTR were only 1.2 and 1.1 mgN/(gVSS·h) and 60.6% and 67.2%. These might be due to the limited EPD activity of denitrifying bacteria at high substrate concentration, when carbon source was not the major determining factor [22].

3.4. Characteristics of Carbon and Nitrogen Transformations via EPD with Different Carbon Sources

Variations in NH₄⁺−N, NO₃⁻−N, NO₂⁻−N, PO₄³⁻−P and COD concentrations of the two reactors in typical cycles on day 195 were analyzed, as shown in Figure 5. At the anaerobic phase (180 min), COD of SBR1^# firstly rapidly decreased from 300 to 77.56 mg/L, and then gradually decreased to 28.4 mg/L, resulting in CRR decreasing from 4.95 to 0.37 mgCOD/(L·min); whereas in SBR2^#, COD gradually decreased from 300 to 29.1 mg/L with an average CRR of 0.95 mgCOD/(L·min). These revealed the faster absorption and transformation of sodium acetate compared to glucose to synthetic intracellular carbons during the anaerobic EPD process [23]. Additionally, NH₄⁺−N and PO₄³⁻-P were kept at 40 and 4 mg/L in both SBRs, with NO₃⁻−N and NO₂⁻−N maintained at low levels.

At the anoxic phase, concentrations of NO₃⁻−N in the SBR1^# and SBR2^# decreased from 70 mg/L to 1.75 and 4.1 mg/L, respectively, with maximum accumulation of NO₂⁻−N observed in 100 and 140 min reaching 57.4 and 60.1 mg/L, respectively; r_NaR values were, respectively, 1.61 and 1.10 mgN/(gVSS·h), and the NO₂⁻−N accumulation rates (r_NiA) were 1.28 and 0.96 mgN/(gVSS·h). Anoxic COD concentrations were almost unchanged in both reactors, demonstrating that partial denitrification was the principal means of removing NO₃⁻−N by consuming the intracellular carbon sources stored at the anaerobic phase. Lately, during the 280~300 min of SBR1^# and 320~340 min of SBR2^#, NO₂⁻−N dropped to 55.0 and 52.6 mg/L, respectively, suggesting that NO₂⁻−N was further denitrified resulting in a slight drop in NO₂⁻−N_eff. The results confirmed once again that precise control of anoxic duration at T_Ni,max was the key to achieve high NO₂⁻−N accumulation [24].

Moreover, at the oxic phase, there were no significant changes in the concentrations of NH₄⁺−N, NO₃⁻−N, NO₂⁻−N, PO₄³⁻−P and COD, indicating that there was no nitration reaction or phosphorus removal process in the system.

3.5. Characteristics of Intracellular Carbons Transformations via EPD with Different Carbon Sources

Variations in PHAs, PHB, PHV and glycogen in SBR1^# and SBR2^# in typical cycles on day 195 are shown in Figure 6. At the anaerobic phase of SBR1^#, glycogen gradually decreased from 20 to 14.15 mmolC/L in 180 min, with a degradation rate of 0.044 mmolC/(L·min); in SBR2^#, glycogen dropped to 13.25 mmolC/L within the first 90 min, with a degradation rate of 0.037 mmolC/(L·min). The difference of glycogen degradation trend was correspondent to the COD removal, confirming that the different carbon sources resulted in internal-to-external carbon transformation [1]. Additionally, it is worth noting that glycogen in SBR2^# increased significantly at 90~180 min, the possible reason being that GAOs could directly synthesize glycogen by using the glucose (40–50 mg/L) presented under condition [25].

According to the variations in PHAs, PHB and PHV concentrations, it could be seen that there was an increase in PHAs of SBR1^# from 14.0 to 22.7 mmolC/L, with a synthesis rate of 0.049 mmolC/(L·min); similarly, PHAs in SBR2^# increased from 12.03 to 19.8 mmolC/L with a synthesis rate of 0.043 mmolC/(L·min) (accounting for about 88% of SBR1^#). The reason was faster-absorbed converted carbon source of sodium acetate. Since glucose has a long carbon chain, it results in fewer GAOs that can be directly transformed and transport it into cells for further utilization before undergoing gene-translocation or degradation and energy generation.

At the anoxic phase, glycogen in SBR1^# and SBR2^# increased from 12.03 and 13.25 mmolC/L to 17.49 and 16.92 mmolC/L, respectively, accompanied by PHAs that decreased from 22.23 and 17.71 mmolC/L to 18.00 and 14.04 mmolC/L. This revealed the respondent intracellular carbon (PHAs) for NO₃⁻−N removal and NO₂⁻−N generation during the EPD process even with a different external carbon source of sodium acetate and glucose. Additionally, the PHV content of SBR2^# was lower than that of SBR1^#, indicating a higher PHB synthesis amount when glucose is used as a carbon source. PHB accounted for approximately 85% of PHAs in SBR2^#, while only 70% in SBR1^#. Thus, variations in intracellular carbons elucidated the different effect of carbon source on EPD performance.

3.6. Analysis of the Variation in Microbial Community under Different Carbon Sources

The microbial communities in sludge samples from SBR1^#and SBR2^# on Day 52 (EPDA-52 and EPDG-52) and Day 195 (EPDA-195 and EPDG-195) were evaluated by high-throughput sequencing analysis at phylum and genus levels, with results presented in Figure 7.

At the phylum level, Proteobacteria, Bacteroidetes, Chloroflexi and Firmicutes predominated (>5%) in SBR1^#and SBR2^#, and Proteobacteria and Bacteroidetes encompass most denitrifying bacteria [26]. Specifically, on Day 52, Proteobacteria was most abundant in SBR1^#and SBR2^#, with 60.68% and 56.80%, respectively. During the EPD system operation, Proteobacteria’s relative abundance decreased in SBR1^# (55.31%) but increased in SBR2^# (62.63%) on Day 195. Concurrently, the relative abundance of Bacteroidetes showed an increasing trend in both reactors; it increased from 10.3% to 12.47% in SBR1^#, and increased from 10.19% to 15.45% in SBR2^#. And there were no significant changes in Chloroflexi and Firmicutes. This may be attributed to differences in carbon sources (sodium acetate and glucose), affecting the microbial community structure in the EPD system [27]. The results indicated that compared to sodium acetate, glucose as the carbon source was better for the proliferation of denitrifying bacteria.

At the genus level, the microbial community analyses revealed the dynamic change of microorganisms in the EPD system, as shown in Figure 7b. On day 52, Candidatus Competibacter was the dominant genus in SBR1^# (EPDA-52) and SBR2^# (EPDG-52), with relative abundances of 49.21% and 48.28%, respectively. And the relative abundance of Defluviicoccus in both reactors was approximately 5%. Previous studies indicates that Candidatus Competibacter and Defluviicoccus contribute to organic matter removal in the partial denitrification process [28,29]. The COD removal efficiency was 83.9% and 77.4% in SBR1^# and SBR2^# on day 52, respectively, proving this conclusion effectively. Meanwhile, the relative abundances of other denitrifying bacteria, such as Rhodocyclaceae and Thauera, were relatively low in both reactors, only about 1.5–3.0%, and had little influence on partial denitrification.

Following 195 days of acclimation and enrichment, Candidatus Competibacter and Defluviicoccus were still the dominant genus in SBR1^# (EPDA-195) and SBR2^# (EPDG-195). However, the relative abundances of Candidatus Competibacter dropped to 16.56% and 18.34%, decreasing by more than 1/2. Conversely, the relative abundance of Defluviicoccus rose to 24.64% and 26.82%, marking a 400% increase and supplanting Candidatus Competibacter as the most abundant genus in the EPD system. These findings indicated that the long-term operation of the EPD system could promote the proliferation of Defluviicoccus. Currently, the COD removal efficiency in SBR1^# and SBR2^# reached 90.5% and 90.3%, respectively. As we know, Defluviicoccus could degrade multiple organic compounds simultaneously [28], which could also explain the COD removal effectively in SBR1^# and SBR2^#.

Additionally, the relative abundance of the denitrifying bacterium (Thauera) increased in both reactors, from 1.94% to 6.43% in SBR1^# and from 3.4% to 12.16% in SBR2^#. Notably, in SBR2^#, the relative abundance of Thauera increased more substantially, suggesting that glucose as a carbon source might promote Thauera proliferation better in the EPD system. And the relative abundance of Rhodocyclaceae also showed a significant upward trend, increasing from 1.96% to 8.21% and 1.53 to 12.56% in the two reactors, respectively. The denitrifying bacteria of Pseudomonas, Comamonas, Rhodobacteraceae and Dechloromonas had no significant changes, with their relative abundances fluctuated by less than 0.5%; additionally, Rhodobacteraceae showed a slight increasing trend, but the other three genera of Pseudomonas, Comamonas and Dechloromonas showed slight decreasing trends (Figure 7b). The increase in the abundance of denitrifying bacterium could effectively improve the denitrification performance of EPD systems. On day 195, the maximum accumulated concentration of NO₂⁻−N reached 57.4 and 60.1 mg/L, respectively, demonstrating a good partial denitrification performance. According to the results, both sodium acetate and glucose as carbon sources could achieve EPD systems successfully. Compared with sodium acetate, glucose as a carbon source was more beneficial to the enrichment of partial denitrifying bacterium, thereby improving the nitrogen removal performance of EPD system.

By the way, there was a very small proportion of AOB (Nitrosomonas) in the EPD system, with a relative abundance of 0.3% at the highest point and only 0.02% at the lowest point. Other bacteria, such as NOB and AnAOB, were almost non-existent; therefore, it could be seen that other bacteria in the reactors have no influence on the performance of EPD system.

4. Conclusions

In this study, characteristics of EPD start-up and optimization were investigated with different carbon sources. By employing two SBRs (SBR1^# and SBR2^#), both successively operated under A/O and A/A/O operation modes, and regulating COD_inf and NO₃⁻−N_inf, EPD could be successfully started up with sodium acetate or glucose as the carbon source.

(1)

During the A/O operation, both SBR1^# and SBR2^# achieved good anaerobic COD removal performances with COD removal efficiency higher than 85% and COD_eff lower than 35 mg/L, but the performance of SBR2^# was significantly affected by high COD (250~350 mg/L). Specially, COD mainly removed in the first 15 min in SBR1^# with CRR reaching 7.54 mgCOD/(L·min), whereas in SBR2^#, COD removal occurred in the whole anaerobic phase (180 min) with an average CRR of 2.22 mgCOD/(L·min).

(2)

By gradually increasing COD_inf (150~250 mg/L) and NO₃⁻−N_inf (20~40 mg/L), both SBR1^# and SBR2^# maintained good partial denitrification performance with high NTR and (r_NaR) of 88.4~90%, 2.41~2.38 mgN/(gVSS·h), respectively, but high COD_inf (250~350 mg/L) and NO₃⁻−N_inf (50~60 mg/L) facilitated stable NO₂⁻−N accumulation in SBR1^# using sodium acetate as the carbon source. Both SBR1^# and SBR2^# reach the maximum NO₂⁻−N accumulation of 54.7 and 62.5 mg/L, respectively, under NO₃⁻−N_inf reaching 70~80 mg/L.

(3)

Using sodium acetate and glucose as carbon sources to drive EPD, similar anaerobic and anoxic internal carbon transformations were observed, but higher and faster carbon transformation was achieved with sodium acetate as carbon source than glucose. Precise control of anoxic time at T_Ni,max was still the key to achieve high NO₂⁻−N accumulation.

(4)

The differences in carbon sources (sodium acetate and glucose) would affect the microbial community structure in the EPD system. Both sodium acetate and glucose as carbon sources could achieve EPD systems successfully; glucose as a carbon source was more beneficial to the enrichment of partial denitrifying bacterium compared to sodium acetate.