Optimization of methane-dependent oxygenic denitrification in sequencing batch reactors by insights into the microbial interactions

Science of the Total Environment 643 (2018) 623–631

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Optimization of methane-dependent oxygenic denitrification in sequencing batch reactors by insights into the microbial interactions Zhanfei He, Jieni Feng, Zhen Wei, Shuyun Wu, Jinte Zou, Xiangliang Pan ⁎ Key Laboratory of Microbial Technology for Industrial Pollution Control of Zhejiang Province, College of Environment, Zhejiang University of Technology, Hangzhou, China

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Decoupling of microbial metabolism and growth was introduced in a new model. • Reactor performance and microbial biomass could be well predicted by the model. • Stimulation of organic matter on methane-dependent O2DN bacteria was discovered. • Great improvements were made by optimizing the operating conditions of SBRs.

a r t i c l e

i n f o

Article history: Received 22 April 2018 Received in revised form 17 June 2018 Accepted 19 June 2018 Available online xxxx Editor: Jay Gan Keywords: Biological nitrogen removal (BNR) N2O emission Methane-dependent O2DN Mathematical modeling Process optimization

a b s t r a c t Methane-dependent oxygenic denitrification (O2DN) is a promising technology used for reducing greenhouse gas emissions of nitrous oxide (N2O) during wastewater treatment. Heterotrophic bacteria are associated with methane-dependent O2DN bacteria, and it has been proposed that metabolic cross-feeding occurs between the two populations above. In this study, a mathematical model was developed to describe the microbial processes and interactions between methane-dependent O2DN bacteria and associated heterotrophic bacteria in a sequencing batch reactor (SBR). A growth factor-dependent decoupling of metabolism and growth of methane-dependent O2DN bacteria was introduced into the model. Effects of influent substrates, operating parameters, and initial biomass on microbial community and reactor performance were then investigated, and the above parameters were optimized using the model. Results surprisingly show that organic matter in the influent greatly stimulated the growth of methane-dependent O2DN bacteria but slightly limited the increase of heterotrophic bacteria. This effect could be explained by the increased excretion of growth factors by heterotrophic bacteria and the intensified competition for nitrite when methane-dependent O2DN bacteria increased. These results will assist in providing a new understanding of microbial interactions in methane-dependent O2DN systems and offer a new and efficient strategy for operating methane-dependent O2DN reactors. © 2018 Elsevier B.V. All rights reserved.

1. Introduction

⁎ Corresponding author at: College of Environment, Zhejiang University of Technology, Hangzhou 310014, China. E-mail address: [email protected] (X. Pan).

https://doi.org/10.1016/j.scitotenv.2018.06.238 0048-9697/© 2018 Elsevier B.V. All rights reserved.

The greenhouse gas nitrous oxide (N2O) is emitted during wastewater treatment, which is a serious environmental issue. A novel biological technology known as oxygenic denitrification (O2DN) was recently developed and can greatly reduce N2O emissions during biological

624

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

nitrogen removal (BNR) processes (reviewed by He et al. (2018)). In the canonical denitrification process, nitrate is reduced to nitrite, nitric oxide (NO), N2O, and nitrogen (N2); however, in the O2DN process, NO is directly disproportionated into N2 and O2, and the intermediate N2O is bypassed (Ettwig et al., 2010). Therefore, emissions of N2O can be avoided during the O2DN process (Ettwig et al., 2009; Shi et al., 2013). Studies have been shown that certain bacteria are capable of reducing nitrate or nitrite into N2 through the novel O2DN pathway when methane or alkanes are employed as electron donors (Zedelius et al., 2011; Ettwig et al., 2012; He et al., 2018). The methane-dependent O2DN process (Eq. (1)) has been intensively studied in the recent years; in this process, methane is completely oxidized into carbon dioxide in an anaerobic environment and nitrite is reduced to N2 via the O2DN pathway. As methane (biogas) is easily available and has a low cost, use of the methane-dependent O2DN process is deemed to be a promising technology for replacing the canonical denitrification process (Luesken et al., 2011a; He et al., 2016). 3CH4 þ 8NO
2 − þ 8Hþ →3CO2 þ 4 N2

−1

þ 10H2 O ΔG00 ¼ −928 kJ  mol

CH4



ð1Þ

Heterotrophic communities are often detected in autotrophic BNR systems (Castro-Barros et al., 2017; Lawson et al., 2017; Reino et al., 2018) and have been found in cultures of methane-dependent O2DN bacteria (He et al., 2015d; Fu et al., 2017). In addition, soluble chemical oxygen demand (COD) was detected in effluent from O2DN reactors (He et al., 2015d), and it is considered to be supplied by the methanedependent O2DN bacteria, as no organic matter (OM) (except methane) was supplied in the influent. These soluble COD can support the growth of heterotrophic bacteria in oligotrophic environments. Methanedependent O2DN bacteria are unable to synthesize certain essential growth factors, such as pyrroloquinoline quinone (PQQ), which is an indispensable cofactor of PQQ-dependent methanol dehydrogenase (MDH) in methane-dependent O2DN bacteria. It has been indicated that PQQ cannot be synthesized by M. oxyfera, a type of methanedependent O2DN bacteria, due to the lack of a pqqABCDE gene cluster (Ettwig et al., 2010; Wu et al., 2015). Therefore, methane-dependent O2DN bacteria will uptake exogenous PQQ produced by other communities. Microbial interactions are complex, and therefore mathematical models are useful tools for revealing internal relationship among different populations (Chen et al., 2015; Winkler et al., 2015; Xu et al., 2016). The present work develops a two-population model with a decoupling of the metabolism and growth of methane-dependent O2DN bacteria to describe the microbial processes and interactions between methane-dependent O2DN bacteria and associated heterotrophic bacteria in a sequencing batch reactor (SBR). Theoretical analysis identifies several key parameters and their effects were investigated through the model. Finally, influent substrates and operating parameters were optimized using the model, and these optimal results were compared with experimental results. 2. Materials and methods 2.1. Model development A two-population model was developed to describe the microbial processes of methane-dependent O2DN bacteria and associated heterotrophic bacteria in an SBR. The setup and operation of the SBR is described in the Supporting Information. The model comprises five solid species and six soluble species: the solid species include active methane-dependent O2DN bacteria (XM), active heterotrophic bacteria (XH), decomposition substance (XS), extracellular polymeric substances (XEPS), and inter biomass (XI); and the soluble species include dissolved methane (SCH4), nitrite (SNO2), readily degradable substrate (SS), utilization-associated products (SUAP), biomass-associated products

(SBAP), and the limiting growth factor of methane-dependent O2DN bacteria (Sgf). Eight microbial processes are involved in the model (see Table 1) and their stoichiometric matrices are shown in Table S1. Fig. 1 illustrates the carbon flows between two populations, with XM and XH biomass cross-fed by exchanging nutrients. XM biomass supplies organic substrates (soluble microbial products, SMP) to the XH biomass and returns growth factors. Experimental findings indicated the existence of SMP, which plays an important role in the methanedependent O2DN system (He et al., 2015d). An unexpected high abundance of heterotrophic denitrifier (~15%) was detected after long-term cultivations in an inorganic medium (He et al., 2013; He et al., 2015c), and it was considered their carbon source was the product of autotrophic microorganisms. Some of the growth factors of methanedependent O2DN bacteria need to be obtained from other microorganisms because they cannot be synthesized by methane-dependent O2DN bacteria (Ettwig et al., 2010; Wu et al., 2015). Therefore, cross-feeding is an important interaction in this system and has been described in the new model. Microbial processes within the SBR are very complicated, and several assumptions were made to simplify the model in this study. For example, some impacts from minor processes and low-yield microorganisms are ignored in the model, such as the impacts of influent, effluent, and settlement processes of the reactor, and low-yield methanogens in the biomass. The decoupling of metabolism and growth of XM had been previously observed (Kampman et al., 2012; He et al., 2015a), and a growth factor-dependent decoupling of metabolism and growth was thus introduced in the model. When Sgf is sufficient, the metabolism of XM is combined with growth; or it stops growing, but metabolism continues. This conversion function based on a logistic function was introduced to describe such growth factor-dependent decoupling of metabolism and growth; the expression is SFgf = 1/[1 + e−A(Sgf−Tgf)], where Tgf is the threshold value of Sgf (mg COD/L), and A is a sufficiently large number. Metabolism and growth are only coupled when Sgf is higher than Tgf. 2.2. Model implementation The SBR was modeled using Aquasim 2.1 software (Reichert, 1998) as the approach described previously (Beun et al., 2001). A virtual system was developed to simulate the operation of the SBR (see Fig. S1), which consisted of three completely mixed compartments. The headspace compartment in Fig. S1 was equivalent to the headspace of the SBR, the SBR compartment was equivalent to the effective volume of the SBR (bulk liquid), and the effluent compartment corresponded to the effluent tank. The volume of the SBR compartment changed over time due to the periodic liquid exchange, whereas the volumes of the other two compartments were fixed. Methane flux was supplied in the headspace compartment. A diffusive link was set between the headspace and the SBR compartments to simulate the diffusion of methane from gas to liquid phase. To simulate the periodic inflow and discharge

Table 1 Microbial processes and their kinetic rates in the model. Microbial process (j)

Kinetic rate expression

1. methane-dependent O2DN bacteria growth 2. methane-dependent O2DN bacteria decay 3. Heterotrophic growth on SS

SNO2 μ M K CH4SCH4 þSCH4 K NO2M þSNO2 K I

K INO2M

NO2M þSNO2

SF gf X M

bMXM SNO2 S μ H;S ηNOx K SSþS I S K NO2H þSNO2 K

K INO2H

NO2H þSNO2

XH

4. Heterotrophic growth on UAP

K INO2H SNO2 μ H;UAP ηNOx K UAPSUAP þSUAP K NO2H þSNO2 K I NO2H þSNO2

XH

5. Heterotrophic growth on BAP

SNO2 μ H;BAP ηNOx K BAPSBAP þSBAP K NO2H þSNO2 K I

K INO2H NO2H þSNO2

XH

6. Heterotrophic bacteria decay 7. XS hydrolysis

b H XH

8. EPS hydrolysis

kh, EPSXEPS

S =X H XH kH K XXþX S =X H

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

kUAPM+sUAP

SCH4

(1)

kEPSM+sEPS

SUAP XEPS )SF XM

(1-kEPSM -kUAPM

(8)

SBAP

gf

(2)

provide growth factor

(6)

XH Scof

625

(4)

feed substrate

XS

(7)

1-kgf

(5)

SS (3)

kgf

Fig. 1. Carbon flow scheme in the two-population model. Arabic numerals in round brackets indicate the microbial processes shown in Table 1. EPS and UAP products from heterotrophic growth on SS are not presented here.

of the SBR, discrete influent and effluent flows were introduced into the SBR and effluent compartments, respectively. An advective link was set between the SBR and effluent compartments. Importantly, a recirculating flow of concentrated sludge was introduced to the SBR compartment to simulate settlement and sludge retention in the SBR. 2.3. Dynamics of XM biomass growth Biological reactor performance is mainly dependent on the growth of the function microorganisms. The growth of the XM biomass is analyzed in this section. Hydraulic retention time (HRT) of an SBR is: HRT ¼

t cycle Vt ¼ t V e =V t V e cycle

ð2Þ

where tcycle is the time of 1 cycle (h), Ve is the exchange volume of liquid (L), and Vt is the valid volume of the SBR (L). The withdrawal rate (rw, 1/h) of XM biomass with effluent can be expressed as: rw ¼

V e X M wX =t cycle V e wX wX ¼ ¼ V tXM V t t cycle HRT

ð3Þ

where wX is the fraction of biomass withdrawal with effluent (mg COD/ mg COD). Based on the mass balance of XM biomass, the apparent specific growth rate of XM biomass (μsM,app, 1/h) in the SBR is given as follows: μ sM;app ¼ r M −bM −r w

ð4Þ

where rM and bM are the specific growth rate and the decay rate of XM biomass, respectively (1/h). According to Table 1 and Table S1, rM can be expressed as: rM ¼ μ M

SCH4 SNO2 K INO2M ð1−kEPSM −kUAPM ÞSF gf K CH4 þ SCH4 K NO2M þ SNO2 K INO2M þ SNO2 ð5Þ

Insertion of the expressions of rM and rw into Eq. (4) gives: μ sM;app ¼ μ M  C M  M NO2M  SF gf −

wX −bM HRT

ð6Þ

where CM is a constant due to the oversupply of methane in the SBR, and MNO2M is a Monod-type equation for nitrite on XM (an equation of variable SNO2), which can be expressed as MNO2M = SNO2/(KNO2M + SNO2)·KINO2M/(KINO2M + SNO2). According to Eq. (6), the concentrations of nitrite and growth factor and the values of operating parameters wX and HRT determine the apparent specific growth rate of XM biomass and their effects were studied in this work. 3. Results and discussion 3.1. Parameter estimation The parameter values involved in the model were derived from previous literature or direct measurements (see Table S2), with the exception of the UAP yield coefficient for XM (kUAPM), the EPS yield coefficient of XM (kEPSM), the yield coefficient for Sgf of XH (kgf), and the threshold value of Sgf for XM growth (Tgf). These four parameters were newly defined in this work and were determined prior to model simulation. Fig. 2a shows the time-course concentration of nitrite in the influent and effluent of the SBR during 138 days of operation, which was used to estimate the above four parameters. The four parameters were given as constant variables in AQUASIM software and were estimated by the best fittings (that is, minimizing the sum of the squares of the weighted deviations between the experimental and simulated results) with the following equation: 2 P experiment χ 2 ¼ ni¼1 ½ðSNO2;i −Ssimulation Þ=σ i experiment  where SNO2,i is the iNO2;i th data point in the experiment or simulation and σi is the standard deviation of the experimental data (Reichert, 1998). The results of parameter estimations showed that kUAPM, kEPSM, kgf, and Tgf were 1.6 × 10−3, 6.5 × 10−3, 2.85 × 10−9 mg COD/mg COD and 1.45 × 10−10 mg COD/L, respectively (see Table S2). The values of kUAPM and kEPSM were largely lower than those of heterotrophic bacteria (see Table S2) (Laspidou and Rittmann, 2002). Effluent nitrite, the nitrogen removal rate (NRR), and the final biomass percentage were predicted by the model using the estimated parameters, and results are shown in Fig. 2a and b. In Fig. 2a, the substantial increase in the amount of nitrite in the influent led to the accumulation of nitrite in the effluent, as the elevated nitrite supply was higher than the nitrite that the biomass could consume. However, with growth of the biomass and an increase in NRR, the elevated influent nitrite was completely removed and effluent nitrite

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

7

Influent (experiment) Effluent (experiment) Effluent (simulation)

6

80

5

60

4 40

3 2

20

1 0

Nitrogen removal rate (mg N/L/d)

100

(a)

0

20

40

(b)

12

60

80

10 8

120

0 140

100

6 4

80

simulation experiment

60 40 20 0

2

100

NRR (experiment) NRR (simulation)

Percentage of biomass (%)

Effluent nitrite (mg N/L)

8

Influent nitrite (mg N/L)

626

0

20

40

60

XM

80

Time (day)

XH

100

120

140

Fig. 2. Experimental and simulation results of influent nitrite, effluent nitrite and NRR of the SBR. (a) Experimental data (symbols) for influent and effluent nitrite and simulation results (lines) for effluent nitrite. (b) Experimental and simulation results of NRR and final community abundance (columns in the inset graph). The sharp decline of NRR on day 84 relates to withdrawal of 0.6 L biomass from the SBR.

gradually decreased to nearly zero. Three large “peaks” were evident in the effluent nitrite profile, due to the increase and subsequent decrease of effluent nitrite during the 138 days of operation, and all of these were successfully predicted by the model (see Fig. 2a). Fig. 2b evidences the efficiency of the model predicting results of NRR. The predicted final percentage of XM biomass was 85.5%, which was very close to realtime quantitative PCR (qPCR) experiments (85 ± 8%) but higher than that from fluorescence micrographs (60–70%) (He et al., 2015d). This result was not unexpected, because all microorganisms were assessed in the fluorescence micrographs with 4′,6′-Diamidino-2-phenylindole (DAPI) stain (He et al., 2015d), whereas only two populations (methane-dependent O2DN bacteria and heterotrophic denitrifiers) were considered in the model and the qPCR results. Moreover, the molar ratio of methane oxidation to nitrite reduction in the SBR was mostly in the range of 2.9–3:8 in the simulation (see Fig. S2); these values were within the experimental range (2.6–3.7:8) (Ettwig et al., 2009; He et al., 2015d) but were slightly lower than the theoretical value of 3:8 (Eq. (1)). The effect of companion heterotrophic bacteria could explain this difference between simulation and theoretical results.

3.2. Effects of influent substrates 3.2.1. Effects of influent nitrite and growth factor XM and XH require nitrite as substrate in the SBR, and both are inhibited by an oversupply of nitrite (Kampman et al., 2012; He et al., 2013; Hartop et al., 2017). Influent nitrite greatly affects the community structure and the performance of the SBR: an insufficient amount of influent nitrite exacerbates competition between the two populations and an excess amount inhibits both. The influent nitrite concentration thus needs to be optimized to provide a high performance. Inhibition of nitrite was introduced in the model by multiplying the kinetic rate expressions by Monod-type inhibition expressions, KINO2M/(KINO2M + SNO2) or KINO2H/(KINO2H + SNO2) (see Table 1). The effects of influent nitrite on biomass growth were studied using the model. Results are presented in Fig. 3. They show that fixation of the influent nitrite concentration was not a good strategy, because nitrite was over supplied in the beginning but was insufficient at the end of the operation. With fixation of influent nitrite, the highest biomass increase of XM was 2.39-fold (obtained at 50 mg N/L of influent nitrite); this was

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

4.0

Relative biomass (times)

3.5 3.0

XM XH

2.5 2.0 1.5 1.0 0.5 0.0

Inf-N30 Inf-N40 Inf-N50 Inf-N60 Inf-N70

real optimized

Fig. 3. Effects of influent nitrite on final biomass levels. Concentrations of influent nitrite were fixed at 30, 40, 50, 60, and 70 mg N/L in groups Inf-N30, Inf-N40, Inf-N50, Inf-N60 and Inf-N70, respectively. Real data of influent nitrite (showed in Fig. 2a) were adopted in the group “real” and optimized concentrations of influent nitrite (showed in Fig. S4) were used in the group “optimized”. The ordinate presents the relative biomass of the final biomass of XM or XH (on day 138) to initial biomass of XM.

much lower than that obtained from the real experiments (3.36-fold). A good strategy was obtained by gradually increasing the amount of nitrite entering the system in accordance with microbial biomass growth, and accurately matching the nitrite supply with its consumption in every cycle of the SBR. Thus, in the “optimized” group (shown in Fig. 2), the concentration of influent nitrite was adjusted so that the concentration of effluent nitrite fell within the range of 0.01 to 0.5 mg N/L in every cycle (see Fig. S3). The nitrite supplied was almost completely consumed in every cycle of the “optimized” group, and nitrite inhibition was as low as possible. As expected, the growth of biomass was improved in the “optimized” group, and the biomass increase of XM was much higher than that obtained in real experiments (3.76 vs. 3.36fold). In addition, there was a negligible percentage change in the XH biomass between the “optimized” and “real” groups (14.8% vs. 14.6%), which means that the N2O emission percentages (percentage of N2O emission in the total amount of nitrogen removal) were similar. According to previous findings, the nitrite concentration within the reactor should be maintained at an optimal amount of approximately 1.92 mmol/L, to obtain the highest activity of methane-dependent O2DN bacteria (He et al., 2013); however, this can only be implemented in continuous flow reactors equipped with expensive nitrite sensors and an automatic control system. The effects of the limiting growth factor level (Sgf) on XM and XH biomasses were investigated and the results are shown in Fig. S4. As expected, the increase in Sgf significantly stimulated growth of XM biomass and greatly elevated its biomass percentage. However, the limiting growth factors for XM growth have not yet been determined, and known information is only that PQQ may be one of its limiting growth factors (Ettwig et al., 2010; Wu et al., 2011). Such ideal strategy cannot be applied in the practical operation of reactors, as the relevant growth factors have not yet been identified. 3.2.2. Effect of influent OM As heterotrophic bacteria are present in the SBR system (He et al., 2015d) and OM is always present in wastewater, the impact of influent OM is assessed in this work. Two strategies used to add influent OM (SS) were adopted in the simulations: a fixed concentration of SS and a fixed ratio of SS/SNO2. Results are presented in Fig. 4. Unexpectedly, it was found that adding small amounts of SS could stimulate the growth of XM. The highest XM biomass increase was obtained at SS influent of 24 mg COD/L and 0.34 mg COD/mg N (with associated 4.21- and 4.38fold increases, respectively). A ratio of 0.34 mg COD/mg N means that approximately 20% of the influent nitrite was reduced by the added

627

OM. Compared to the case where no SS was added (abscissa is zero in Fig. 4), there were improvements in the XM biomass of 25.3% and 30.4% with SS additions of 24 mg COD/L and 0.34 mg COD/mg N, respectively. However, more unexpectedly, the addition of small amounts of SS slightly decreased the final biomass of heterotrophic bacteria, and the lowest values were obtained at an influent SS of 25 mg COD/L and 0.33 mg COD/mg N, representing a decrease of 20.5% and 20.7%, respectively, relative to the simulation without influent SS. As expected, when the addition of influent SS was higher than optimal values (above 24–25 mg COD/L and 0.33–0.34 mg COD/mg N), there was a rapid increase in the XH biomass, and a rapid decrease of the XM biomass with increasing amounts of SS. Furthermore, the highest XM biomass percentage was found at 25 mg COD/L and 0.33 mg COD/mg N, with values of 90.3% and 90.6%, respectively. These results are inconsistent with the common perception that OM stimulates the growth of XH. The stimulation of autotrophic microbial communities by OM has also been observed in previous studies and has mostly been attributed to the role of organic metabolites (Lehtovirta-Morley et al., 2014; Antony et al., 2017). To explore the reasons for OM stimulating XM growth, average values for SFgf (switch function of growth factor) and MNO2M (Monod-type equation of nitrite on XM) were calculated in Fig. S5. The results showed that SFgf increased and MNO2M decreased with an increase in influent SS. The increase in SFgf value was attributed to the increase in Sgf excreted by XH in the SBR. Therefore, the addition of influent SS stimulated the production of growth factors by XH, and then indirectly stimulated XM growth (see Fig. S6). However, when the amount of influent SS exceeded the optimal level (0.34 mg COD/mg N) in Fig. S5, there was no further increase in the average value of SFgf (the value was very close to one) but that of MNO2M decreased continuously. This result could then explain the reduction in XM biomass when SS is over supplied. To explore the reason for the decrease in XH biomass when only small amounts of influent SS were added, the reaction system was then simulated by adding extra nitrite to the influent (see Fig. S7). The results show that with the addition of 21 mg N/L of nitrite, the XH biomass completely recovered to levels when influent SS was not added, which indicated that added OM did not inhibit XH (as expected) but rather stimulated the growth of XM, thereby exacerbating nitrite competition. XH biomass reduction can thus be attributed to the intensified competition for nitrite, which has a weaker affinity for nitrite than XM due to the higher affinity constant value (see Table S3). When sufficient nitrite was supplied, nitrite competition weakened and the XH biomass recovered. 3.3. Effects of operating parameters 3.3.1. Effect of withdrawal fraction The SBR is proficient at retaining biomass and is suitable for slowgrowing bacteria (Strous et al., 1998); however, not all biomass is retained, regardless of the settling time. In previous works, XM bacteria were detected in the effluent tanks of XM incubators (Luesken et al., 2011b), and a large proportion of XM biomass was withdrawn with effluent (Kampman et al., 2012; He et al., 2013). The withdrawn fraction (wX) is an important parameter that affects the solid species in the SBR, and its effect on biomass was studied here (results are shown in Fig. 5a). As expected, biomass increased with a decrease in wX, but XH biomass increased discontinuously when wX was below 0.02. Reducing wX would thus be an ideal strategy, as it not only greatly improves the abundance of XM biomass but also its biomass percentage; this would then increase NRR but reduce the N2O emission percentage. If wX was zero (for example, if the membrane module was equipped for liquid exchange), the increase in XM biomass could reach 4.06-fold (21.0% improvement) after 138 days and its percentage would be increased to 87.9%. Indeed, it has been demonstrated that membrane bioreactors could significantly accelerate the cultivation of slow-growing microorganisms, such as anaerobic ammonium oxidation (anammox) bacteria and anaerobic methanotrophic microorganisms (Van Der Star et al.,

628

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

XM XH

4

80

3

60

2

40

1 0

20

Percentage of XM 0

5

20

40

60

80

100

Influent SS (mg COD/L)

120

0

140

100

(b)

4

80

3

60

2

40

1

20

0

Percentage of XM (%)

Relative biomass (times)

100

(a)

Percentage of XM (%)

Relative biomass (times)

5

0.0

0.5

1.0

1.5

2.0

Influent SS/SNO2 (mg COD/mg N)

2.5

0

Fig. 4. Effect of influent OM on final biomass levels and percentages. (a) Fixed concentration of influent OM (SS) was adopted in every simulation. (b) Influent OM increased with an increase in influent nitrite with fixed ratio of SS/SNO2.

2008; Kampman et al., 2014; He et al., 2018). When wX exceeded 0.052, the XM biomass increase was lower than 1.0: the withdrawal rate of the XM biomass was higher than the growth rate, which should be avoided in practical applications. The percentage of XM biomass increased significantly when wX exceeded 0.02 (the proportion of XM biomass reached 98.3% when wX was 0.1), whereas the XH biomass decreased rapidly (faster than the XM biomass) as wX increased. The high withdrawal rates and low SMP levels produced by the XM biomass could explain this rapid decline. 3.3.2. Effect of HRT HRT is a key parameter of reactor operation and it affects not only soluble but also solid species. High HRT means a low flow of liquid exchange, and a subsequently low biomass withdrawal rate. In this study, simulations were performed using different HRT values but with fixed nitrite loading; results are presented in Fig. 5b. For simplicity, the period of the SBR cycle was fixed (3 days), and the volume of liquid exchange in each cycle was altered to achieve a given amount of HRT. The results indicate that an HRT increase was conducive to the growth of XM biomass. With an increase in HRT over a short period, there was a rapid increase in the growth of XM biomass; however, the increase was slow when HRT was increased over a period exceeding 30 days.

According to the simulations, the increase in XM biomass was 4.55fold at an HRT of 90 days, which was 35.5% higher than that at 7.5 days (the real condition of the SBR). However, the increase in XH biomass was negligible when HRT exceeded 7.5 days, but the percentage of XM biomass increased significantly (from 85.5% at HRT of 7.5 days to 88.6% at HRT of 90 days). These results imply that increasing HRT is a good method for enhancing reactor performance while limiting N2O emissions from the SBR. Long HRTs (2.0–65 days) have been adopted in previous cultivations of methane-dependent O2DN bacteria (Raghoebarsing et al., 2006; Ettwig et al., 2009; Luesken et al., 2011b; Hu et al., 2014; He et al., 2015b). However, an opposite result was found in a membrane bioreactor, where NRR increased with a decrease in HRT from 61 to 1.4 days. This result was explained by the mitigation of the possible inhibition by unidentified intermediates or products at shorter HRTs (Kampman et al., 2014). The possible inhibition by intermediates or products deserves consideration in the operation of reactors, but further evidence is required to verify this finding. 3.4. Effect of initial biomass Inoculum (initial biomass) is important for the enrichment of methane-dependent O2DN bacteria (He et al., 2015a). The impact of

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

95

3

XM XH Percentage of XM

2 1

5

0.02

0.04

wX

0.06

90

85

80 0.10

0.08

92

(b)

Percentage of XM (%)

4

90

3

Growth of XM biomass (times)

3.5

4

0 0.00

Relative biomass (times)

4.0

100

(a)

Percentage of XM (%)

Relative biomass (times)

5

629

3.0

L D/ /L CO L g OD /L D/ m C /L 00 CO OD mg C g OD 5 =1 g m 2 C m = XH =5 mg XH =1 .2 XH XH =0 XH

2.5 2.0 1.5 1.0 0.5 0.0

0

5

10

15

XM biomass (mg COD/L)

20

25

Fig. 6. Effect of initial biomass on growth of XM biomass. Initial XH biomass is identical on a certain curve and its value is labeled near the curve.

88 2

86

1 0

0

20

40

HRT (day)

60

80

84

Fig. 5. Effects of withdrawal fraction (a) and HRT (b) on final biomass levels and percentages. The real conditions in the SBR were a withdrawal fraction of 0.02 and a HRT of 7.5 days. Nitrite loading was fixed for all simulations with different HRTs.

the initial XM and XH biomasses on XM biomass growth during 138 days of operation were investigated, and results are shown in Fig. 6. Overall, the growth of XM biomass was positively correlated with the initial biomass of both XM and XH, although that of XM was more sensitive than XH, especially at high levels. With an increase in the initial XM biomass, the initial XH biomass had an increased effect on XM biomass growth. For example, when the initial XH biomass increased from 25 to 100 mg COD/L, the XM biomass growth increased 0.31-fold at an initial XM biomass of 3 mg COD/L, and 0.93-fold at 15 mg COD/L. However, some combinations using low initial biomass levels resulted in a decrease in the XM biomass during the 138 days operation (ordinate below 1.0 in Fig. 6), indicating operational failure. According to Eq. (6), the specific growth rate of XM biomass was apparently independent of the absolute level of XM biomass, but it was actually dependent (see Fig. 6). The initial biomass, neither XM nor XH, might supply OM (SMP or decay products) to XH to produce growth factors (increasing the value of SFgf in Eq. (6)), and then stimulated the growth of XM biomass. To confirm the above explanation, exogenous OM was added in one failed combination of initial biomass (see Fig. S8). The results showed that the addition of SS changed the tendency of the time-course XM biomass in the SBR. The XM biomass increased by 3.70-fold at 20 mg COD/L of influent SS, but it decreased by 51.3% without influent SS. The added SS fed XH to produce growth factors, which indirectly stimulated growth of the XM biomass. Hence, the addition of appropriate amounts of SS in influent is a good strategy, particularly when the initial biomass is low. It was also found that with constant operating conditions, the final communities of XM and XH were extremely steady and independent of the initial biomass (see Fig. S9). Six different combinations of initial XM and XH biomass were adopted in model simulations, and the final combination in all six simulations tended to reach a stable point. This finding indicated that the communities of XM and XH were steady in the reactor system

under constant conditions and that the initial biomass did not determine the final steady state of the community. 3.5. Optimization of operating conditions As demonstrated above, influent substrates and operating parameters greatly affect XM biomass abundance and performance of the SBR. A combination of optimized conditions was thus designed based on the above findings. This included an optimized influent nitrite concentration, 0.34 mg COD/mg N influent SS, and 30 days HRT. Influent nitrite was optimized to reduce effluent nitrite to the range of 0.01 to 0.5 mg N/L as described above, and the optimized results are shown in Fig. S10. The HRT was set at 30 days as only a slight improvement (approximately 6%) was achieved between 30 days and 90 days or longer, but the risk of intermediate or product inhibition increased greatly. Configuration of the SBR was not modified, and there was thus no change in the optimization of wX. A comparison of NRR, XM biomass increase, and XM biomass percentage under optimized conditions with the results obtained from experiments (He et al., 2015d) is shown in Fig. 7. Compared with real data, NRR rose considerably under optimized conditions from 12.3 to 31.7 mg N/L/day, achieving an improvement of 158.2%; XM biomass increased from 3.36 to 10.0-fold, and the XM biomass percentage increased from 85.5% to 87.6% (the XH biomass percentage decreased from 14.5% to 12.4%). These results are ideal, as the reactor performance was greatly improved and the potential N2O emission percentage was reduced. The changes in biomass and substrate during 1 cycle of the SBR require investigation with respect to exploring the optimization mechanism (see Fig. S11). Two phases were distinguished in 1 cycle under the optimized conditions: an XH growth phase and an XH decay phase. In the XH growth phase (the first 5 h in the final cycle), the XH biomass increased quickly and was accompanied by accumulation of Sgf and the rapid consumption of nitrite and SS. In the XH decay phase (hours 5–72), SS was nearly depleted, which resulted in a decrease in the XH biomass. Nitrite was reduced continuously in the XH decay phase, but the rate dropped with a decrease in the XH biomass. At the end of the experiment, the nitrite concentration was near zero, which decreased the average concentration of the nitrite in SBR and relieved the substrate inhibition of nitrite; Sgf then returned to low levels due to consumption of XM biomass growth. These two phases alternated during SBR operation under optimized conditions. The XM biomass grew exponentially in both phases due to sufficient growth factors being produced by the XH biomass in the XH growth phase, which resulted in the rapid increase of the XM biomass under optimized conditions.

630

Z. He et al. / Science of the Total Environment 643 (2018) 623–631

25 20 15

87

8

Percentage of XM (%)

30

88

real conditions optimized

10

XM biomass increase (times)

Nitrogen removal rate (mg N/L/d)

35

6

86

4

85

2 0

84

XM increase

Percentage of XM

20

40

optimized results real data

10 5 0

0

60

80

100

120

140

Time (day) Fig. 7. Comparison between the optimized results and the real data in the SBR. Inset graph shows biomass increases of XM and final XM biomass percentages.

4. Conclusion A two-population model was established in which microbial interactions between bidirectional cross-feeding of methane-dependent O2DN bacteria and heterotrophic bacteria were described, and a growth factor-dependent decoupling of metabolism and growth of methanedependent O2DN bacteria was introduced. The influent substrates, operating parameters, and initial biomass were found to primarily affect the microbial community and the reactor performance of the SBR studied. Influent OM was unexpectedly found to significantly stimulate the growth of methane-dependent O2DN bacteria but to slightly decrease the heterotrophic biomass, thereby resulting in high levels and percentages of methane-dependent O2DN bacteria. Exogenous OM stimulated the excretion of growth factor by heterotrophic bacteria and the growth of methane-dependent O2DN bacteria, whereas it increased the biomass of methane-dependent O2DN bacteria and intensified the competition for nitrite, which caused a slight decrease in the heterotrophic biomass. Based on these findings, an optimized strategy was proposed for operating the SBR with a high NRR (improvement of 158.2%) and achieving a low N2O emission percentage (decrease of heterotrophic biomass percentage from 14.5% to 12.4%). Acknowledgements This work was funded by the National Natural Science Foundation of China (41701274, U1503281 and U1403181). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.scitotenv.2018.06.238. References Antony, R., Willoughby, A.S., Grannas, A.M., Catanzano, V., Sleighter, R.L., Thamban, M., Hatcher, P.G., Nair, S., 2017. Molecular insights on dissolved organic matter transformation by supraglacial microbial communities. Environmental Science & Technology 51 (8), 4328–4337.

Beun, J.J., Heijnen, J.J., Van Loosdrecht, M.C.M., 2001. N-removal in a granular sludge sequencing batch airlift reactor. Biotechnology and Bioengineering 75 (1), 82–92. Castro-Barros, C.M., Jia, M., van Loosdrecht, M.C.M., Volcke, E.I.P., Winkler, M.K.H., 2017. Evaluating the potential for dissimilatory nitrate reduction by anammox bacteria for municipal wastewater treatment. Bioresource Technology 233, 363–372. Chen, X., Guo, J., Xie, G.J., Liu, Y., Yuan, Z., Ni, B.J., 2015. A new approach to simultaneous ammonium and dissolved methane removal from anaerobic digestion liquor: a model-based investigation of feasibility. Water Research 85, 295–303. Ettwig, K.F., van Alen, T., van de Pas-Schoonen, K.T., Jetten, M.S., Strous, M., 2009. Enrichment and molecular detection of denitrifying methanotrophic bacteria of the NC10 phylum. Applied and Environmental Microbiology 75 (11), 3656–3662. Ettwig, K.F., Butler, M.K., Le Paslier, D., Pelletier, E., Mangenot, S., Kuypers, M.M., Schreiber, F., Dutilh, B.E., Zedelius, J., de Beer, D., Gloerich, J., Wessels, H.J., van Alen, T., Luesken, F., Wu, M.L., van de Pas-Schoonen, K.T., Op den Camp, H.J., Janssen-Megens, E.M., Francoijs, K.J., Stunnenberg, H., Weissenbach, J., Jetten, M.S., Strous, M., 2010. Nitrite-driven anaerobic methane oxidation by oxygenic bacteria. Nature 464 (7288), 543–548. Ettwig, K.F., Speth, D.R., Reimann, J., Wu, M.L., Jetten, M.S., Keltjens, J.T., 2012. Bacterial oxygen production in the dark. Frontiers in Microbiology 3, 273. Fu, L., Ding, J., Lu, Y.Z., Ding, Z.W., Zeng, R.J., 2017. Nitrogen source effects on the denitrifying anaerobic methane oxidation culture and anaerobic ammonium oxidation bacteria enrichment process. Applied Microbiology and Biotechnology 101 (9), 3895–3906. Hartop, K.R., Sullivan, M.J., Giannopoulos, G., Gates, A.J., Bond, P.L., Yuan, Z., Richardson, D.J., 2017. The metabolic impact of extracellular nitrite on aerobic metabolism of Paracoccus denitrificans. Water Research 113, 207–214. He, Z., Cai, C., Geng, S., Lou, L., Xu, X., Zheng, P., Hu, B., 2013. Modeling a nitrite-dependent anaerobic methane oxidation process: parameters identification and model evaluation. Bioresource Technology 147, 315–320. He, Z., Cai, C., Shen, L., Lou, L., Zheng, P., Xu, X., Hu, B., 2015a. Effect of inoculum sources on the enrichment of nitrite-dependent anaerobic methane-oxidizing bacteria. Applied Microbiology and Biotechnology 99 (2), 939–946. He, Z., Geng, S., Cai, C., Liu, S., Liu, Y., Pan, Y., Lou, L., Zheng, P., Xu, X., Hu, B., 2015b. Anaerobic oxidation of methane coupled to nitrite reduction by halophilic marine NC10 Bacteria. Applied and Environmental Microbiology 81 (16), 5538–5545. He, Z., Geng, S., Shen, L., Lou, L., Zheng, P., Xu, X., Hu, B., 2015c. The short- and long-term effects of environmental conditions on anaerobic methane oxidation coupled to nitrite reduction. Water Research 68, 554–562. He, Z., Wang, J., Zhang, X., Cai, C., Geng, S., Zheng, P., Xu, X., Hu, B., 2015d. Nitrogen removal from wastewater by anaerobic methane-driven denitrification in a lab-scale reactor: heterotrophic denitrifiers associated with denitrifying methanotrophs. Applied Microbiology and Biotechnology 99 (24), 10853–10860. He, Z., Geng, S., Wang, L., Cai, C., Wang, J., Liu, J., Zheng, P., Xu, X., Hu, B., 2016. Improvement of mineral nutrient concentrations and pH control for the nitrite-dependent anaerobic methane oxidation process. Separation and Purification Technology 162, 148–153. He, Z., Feng, Y., Zhang, S., Wang, X., Wu, S., Pan, X., 2018. Oxygenic denitrification for nitrogen removal with less greenhouse gas emissions: microbiology and potential applications. Sci. Total Environ. 621, 453–464.

Z. He et al. / Science of the Total Environment 643 (2018) 623–631 Hu, B., He, Z., Geng, S., Cai, C., Lou, L., Zheng, P., Xu, X., 2014. Cultivation of nitritedependent anaerobic methane-oxidizing bacteria: impact of reactor configuration. Applied Microbiology and Biotechnology 98 (18), 7983–7991. Kampman, C., Hendrickx, T.L., Luesken, F.A., van Alen, T.A., Op den Camp, H.J., Jetten, M.S., Zeeman, G., Buisman, C.J., Temmink, H., 2012. Enrichment of denitrifying methanotrophic bacteria for application after direct low-temperature anaerobic sewage treatment. Journal of Hazardous Materials 227-228, 164–171. Kampman, C., Temmink, H., Hendrickx, T.L., Zeeman, G., Buisman, C.J., 2014. Enrichment of denitrifying methanotrophic bacteria from municipal wastewater sludge in a membrane bioreactor at 20 degrees. Journal of Hazardous Materials 274C, 428–435. Laspidou, C.S., Rittmann, B.E., 2002. Non-steady state modeling of extracellular polymeric substances, soluble microbial products, and active and inert biomass. Water Research 36 (8), 1983–1992. Lawson, C.E., Wu, S., Bhattacharjee, A.S., Hamilton, J.J., McMahon, K.D., Goel, R., Noguera, D.R., 2017. Metabolic network analysis reveals microbial community interactions in anammox granules. Nature Communications 8, 15416. Lehtovirta-Morley, L.E., Ge, C., Ross, J., Yao, H., Nicol, G.W., Prosser, J.I., 2014. Characterisation of terrestrial acidophilic archaeal ammonia oxidisers and their inhibition and stimulation by organic compounds. FEMS Microbiology Ecology 89 (3), 542–552. Luesken, F.A., Sanchez, J., van Alen, T.A., Sanabria, J., Op den Camp, H.J., Jetten, M.S., Kartal, B., 2011a. Simultaneous nitrite-dependent anaerobic methane and ammonium oxidation processes. Applied and Environmental Microbiology 77 (19), 6802–6807. Luesken, F.A., van Alen, T.A., van der Biezen, E., Frijters, C., Toonen, G., Kampman, C., Hendrickx, T.L., Zeeman, G., Temmink, H., Strous, M., Op den Camp, H.J., Jetten, M.S., 2011b. Diversity and enrichment of nitrite-dependent anaerobic methane oxidizing bacteria from wastewater sludge. Applied Microbiology and Biotechnology 92 (4), 845–854. Raghoebarsing, A.A., Pol, A., van de Pas-Schoonen, K.T., Smolders, A.J., Ettwig, K.F., Rijpstra, W.I., Schouten, S., Damste, J.S., Op den Camp, H.J., Jetten, M.S., Strous, M., 2006. A microbial consortium couples anaerobic methane oxidation to denitrification. Nature 440 (7086), 918–921. Reichert, P., 1998. AQUASIM 2.0-Computer Program for the Identification and Simulation of Aquatic Systems. Version 2.0EAWAG, Dubendorf, Switzerland.

631

Reino, C., Suárez-Ojeda, M.E., Pérez, J., Carrera, J., 2018. Stable long-term operation of an upflow anammox sludge bed reactor at mainstream conditions. Water Research 128(, 331–340. Shi, Y., Hu, S., Lou, J., Lu, P., Keller, J., Yuan, Z., 2013. Nitrogen removal from wastewater by coupling anammox and methane-dependent denitrification in a membrane biofilm reactor. Environmental Science & Technology 47 (20), 11577–11583. Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The sequencing batch reactor as a powerful tool for the study of slowly growing anaerobic ammonium-oxidizing microorganisms. Applied Microbiology and Biotechnology 50 (5), 589–596. Van Der Star, W.R., Miclea, A.I., Van Dongen, U.G., Muyzer, G., Picioreanu, C., van Loosdrecht, M., 2008. The membrane bioreactor: a novel tool to grow anammox bacteria as free cells. Biotechnology and Bioengineering 101 (2), 286–294. Winkler, M.K., Ettwig, K.F., Vannecke, T.P., Stultiens, K., Bogdan, A., Kartal, B., Volcke, E.I., 2015. Modelling simultaneous anaerobic methane and ammonium removal in a granular sludge reactor. Water Research 73, 323–331. Wu, M.L., Ettwig, K.F., Jetten, M.S., Strous, M., Keltjens, J.T., van Niftrik, L., 2011. A new intra-aerobic metabolism in the nitrite-dependent anaerobic methane-oxidizing bacterium ‘Candidatus Methylomirabilis oxyfera’. Biochemical Society Transactions 39 (1), 243–248. Wu, M.L., Wessels, H.J., Pol, A., Op den Camp, H.J., Jetten, M.S., van Niftrik, L., Keltjens, J.T., 2015. An XoxF-type methanol dehydrogenase from the anaerobic Methanotroph ‘Candidatus Methylomirabilis oxyfera’. Applied and Environmental Microbiology 81 (4), 1442–1451. Xu, G., Yin, F., Chen, S., Xu, Y., Yu, H.Q., 2016. Mathematical modeling of autotrophic denitrification (AD) process with sulphide as electron donor. Water Research 91, 225–234. Zedelius, J., Rabus, R., Grundmann, O., Werner, I., Brodkorb, D., Schreiber, F., Ehrenreich, P., Behrends, A., Wilkes, H., Kube, M., Reinhardt, R., Widdel, F., 2011. Alkane degradation under anoxic conditions by a nitrate-reducing bacterium with possible involvement of the electron acceptor in substrate activation. Environmental Microbiology Reports 3 (1), 125–135.