Application of dispersed and immobilized hydrolases for membrane fouling mitigation in anaerobic membrane bioreactors

Journal of Membrane Science 491 (2015) 99–109

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Application of dispersed and immobilized hydrolases for membrane fouling mitigation in anaerobic membrane bioreactors Philip Chuen Yung Wong a,b,n, Jia Yi Lee a, Chee Wee Teo a,b a

School of Civil and Environmental Engineering, Nanyang Technological University, Singapore 639798, Singapore Singapore Membrane Technology Center, Nanyang Environment and Water Research Institute, Nanyang Technological University, 1 Cleantech Loop, Singapore 637141, Singapore

b

art ic l e i nf o

a b s t r a c t

Article history: Received 11 December 2014 Received in revised form 14 May 2015 Accepted 16 May 2015 Available online 23 May 2015

Enhancing the hydrolysis of microbial macromolecules by supplementing exogenous hydrolases may improve membrane performance via structural disruptions of fouling layers and alterations to sludge characteristics. This was investigated by short batch filtration ( o 1 h) and 30-day extended filtration experiments using laboratory-scale anaerobic membrane bioreactors. Crude hydrolases were either dispersed directly into the reactor, or immobilized onto microfiltration membranes. Under constant flux operation, dispersed enzymes consistently moderated increases in transmembrane pressures (TMPs) compared to the control setup. Immobilized hydrolases appeared effective in the short filtration test, but in the extended experiment, the pseudo-stable TMP was not significantly lower compared to the control TMP. With dispersed enzymes, the average TMP was almost 30% lower than the control value. This was associated with a 33% reduction in the protein content of the bulk extracellular polymeric substances, and a 45% reduction in the membrane cake density. Immobilized enzymes limited cake formation to a similar extent through hydrolysis at the base of the cake, but this was negated by the increase in gel resistance attributed to the hydrophobic attraction between the immobilization layer and proteinaceous hydrolysis products. Even as the dispersed hydrolases exhibited greater effectiveness under the conditions studied, there is scope for further enhancement in both approaches. & 2015 Elsevier B.V. All rights reserved.

Keywords: Submerged anaerobic membrane bioreactor Extracellular polymeric substances Enzyme augmentation Hydrolysis Enzyme-immobilized membrane

1. Introduction The recent focus on energy conservation in sewage treatment [1] has directed a trend toward anaerobic processes, including the employment of anaerobic membrane bioreactors (AnMBRs). Although AnMBRs possess inherent advantages such as the elimination of outlays associated with oxygen provision and low sludge production, membrane fouling control and membrane cleaning still incur sizable expenses. For instance, approximately 47% of the operating cost of AnMBR is associated with gas scouring for fouling mitigation [2]. Several complementary or adjunctive fouling control strategies include supplementing with additives such as powdered and granular activated carbon [3,4], coagulants [5], and quorum quenchers [6]. These additives differ in their mechanistic functionalities. Powdered activated carbon adsorbs macromolecular and colloidal foulants [4], while granular activated carbon has additional membrane scouring ability [3]. Coagulants aggregate colloidal foulants and promote bioflocculation, hence n Corresponding author at: School of Civil and Environmental Engineering, Nanyang Technological University, Singapore 639798, Singapore. Tel.: þ 65 6790 5269; fax: þ 65 6791 0676. E-mail address: [email protected] (P.C.Y. Wong).

http://dx.doi.org/10.1016/j.memsci.2015.05.022 0376-7388/& 2015 Elsevier B.V. All rights reserved.

reducing internal fouling and specific cake resistance [5,7]. Quorum quenchers degrade microbially produced autoinducers to inhibit intercellular communication during biofilm formation, thereby mitigating biofouling. One common characteristic amongst these additives is their passive participation in the removal of microbial macromolecules, which are one of the most significant contributors toward membrane fouling in membrane bioreactors [8,9]. Extracellular polymeric substances (EPS) and soluble microbial products (SMP) are largely constituted by microbial macromolecules. The former are heterogeneous polymeric matrices which envelope and confer adhesive properties to microbial cells [10,11], while the latter are by-products excreted during endogenous decay and substrate utilization [12]. Both comprise principally of proteins, carbohydrates, glycoproteins, and small quantities of lipids, humic substances, nucleic acids and uronic acids [13]. A different fouling control strategy involves utilizing enzymes with affinities toward macromolecules, thereby promoting their degradation. This is the basis behind membrane cleaning agents containing protease, amylase, cellulase and/or peroxidase [14,15], and underpins the supplementation of crude hydrolases to bioreactors to enhance solids hydrolysis [16,17]. The latter may be expedient for AnMBRs treating wastewaters containing particulates such as

100

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

raw sewage, owing to net solids accumulation when the influent particulates are not promptly hydrolyzed [17,18]. Therefore, the dual benefits of mitigating membrane fouling and limiting solids buildup provide the rationale for an enzymatic approach. Hydrolases can potentially alleviate membrane fouling by reducing the quantities, or through structural simplifications of macromolecular EPS and SMP. This is partially supported by demonstrations of the affinity of these enzymes towards EPS in sludge digestion [19,20] and biofilm degradation [21]. The benefits of supplementing dispersed or membrane-immobilized hydrolases on membrane filtration have also been reported. Hodgson et al. [22] ascribed the 2–3 fold reduction in filtration resistance following the addition of 100 mg/mL protease to a carbon-limited bacterial suspension to the enzymatic modification of EPS matrix. Howell and Velicangil [23] filtered 0.5% albumin and 0.5% hemoglobin separately using ultrafiltration membranes with 0.2–0.5% attached protease. Membrane flux were 25–78% higher compared to that of the unmodified membranes. Presently, the costs of crude protease and amylase produced via submerged fermentation are around US$5/kg [24] and US$10/kg [25], respectively, which do not make their use overtly favorable. However, such costs are likely to be substantially moderated when solid-state fermentation of reusable substrates such as agroindustrial wastes [26,27] and wastewater sludge [28] becomes more widespread. For example, the manufacturing cost of cellulase can be reduced from US$20/kg to US$0.20/kg by switching from submerged to solid-state fermentation [27]. A similar extent in cost reduction for protease and amylase will improve the economic feasibility of the proposed strategy. The outlook is further enhanced by continued advancements in protein engineering and molecular farming techniques [29]. The objective of this research was to investigate the effects of enzyme augmentation on filtration performances of AnMBRs treating synthetic sewage. Supplementation in the form of dispersed and membrane-immobilized hydrolases was examined and contrasted. Batch filtration tests were performed under supracritical flux, while protracted continuous filtration was conducted under a moderate flux to assess the effectiveness of these hydrolases as fouling control agents. Apart from assessing filterability using transmembrane pressure (TMP), the compositions of membrane foulants were analyzed to deduce fouling mitigation mechanisms.

2. Materials and methods 2.1. Laboratory scale AnMBRs Batch and extended filtration experiments were performed using two submerged AnMBRs with working volumes of 0.7 L and 5 L, respectively. The schematic of the latter is shown in Fig. 1. The setup for the smaller reactor was similar except for the absence of computer control and feeding apparatus. For both AnMBRs, headspace biogas was recirculated at 1.5 L/min to provide adequate mixing and to reduce solids deposition on the membrane surfaces. The gas flow translated to superficial velocities of 20 m/h and 4.7 m/h for the smaller and larger bioreactors, respectively. Flat-sheet polyvinylidene fluoride membranes (HVLP, Millipore, USA) with a nominal pore size of 0.45 mm were installed in two custom-made membrane modules (Ying Kwang, Singapore) with effective membrane areas of 196 cm2 and 0.05 m2. These membranes were pre-wetted and washed with distilled water to remove any antimicrobial preservatives before use. Seed sludge was obtained from an anaerobic digester treating municipal sludge (Ulu Pandan Water Reclamation Plant, Singapore). The 5 L AnMBR was fed with synthetic wastewater

constituted by twice diluted standard sewage (OECD 303 A), and 250 mg/L of dog food (ALPO, Purina, USA) grinded to less than 2 mm. ALPO dog food was used as the particulate organic component because of its compositional similarity with primary sludge [30]. The soluble and particulate chemical oxygen demand (COD) of the feed were 150 mg/L and 400 750 mg/L, respectively, giving a total COD of 550 750 mg/L. Sludge employed in the batch experiments was cultivated for one month with the same feed in a fed-batch manner in a 2 L carboy (1.12 kg COD/m3/d volumetric loading rate,  20 g/L reactor total suspended solids (TSS), pH 77 1) without wasting. 2.1.1. Batch filtration The final feed for the batch-cultivated sludge was provided around 4 h before the experiment. Reactor TSS was diluted to 2500 mg/L using tap water, and pH adjusted to 7.070.2 using HCl and NaOH (5 N each). Each set of batch experiment comprised of a “test” and “control” pair with and without enzyme supplementation, respectively. Filtration in the control run was terminated when the TMP reached approximately 40 kPa. The test run was then conducted for at least the same elapsed time. Flowrate and TMP were measured by a variable area rotameter (FR2L09BVBN, Key Instruments, USA) and a digital pressure gauge (DPG1000B, Cecomp Electronics, USA), respectively. Tap water was added manually via an external port to maintain a constant liquid level. Purified enzymes at doses of 35 and 14 mL (equivalent to 90 and 36 mL/g feed COD) were dispersed in the sludge suspension for the first two experimental sets prior to filtration. The corresponding operating flux were 40 and 20 LMH for the higher and lower doses, respectively. The third experimental set using membraneimmobilized hydrolases was performed at 40 LMH. These supracritical flux values were selected to observe accelerated fouling behaviors over short time periods. Critical flux as determined by flux-stepping 1 LMH every 10 min [31], until the rate of TMP increase exceeded 0.05 kPa/min, were approximately 13 and 19 LMH for the control and enzyme-supplemented runs, respectively. The dispersed enzyme experiments were conducted on consecutive days and the immobilized enzyme experiment 3 weeks thereafter. The extent of fouling mitigation was quantified by the percentage reduction in TMP, ΔTMP, and a temporal retardation factor, t t =t c . The former is defined as   TMP c  TMP t  100%; ð1Þ ΔTMP ¼ TMP c where TMP c and TMP t represent the time-varying TMP values of the control and test bioreactors at the same instant, respectively. In the latter scale factor, t c denotes the time taken for the control setup to reach a certain TMP c , and t t symbolizes the corresponding time taken for the test reactor to reach the same TMP. t t =t c and ΔTMP allow assessments on the relative effectiveness of the augmentation strategies by reducing the influences of nuisance factors and lurking variables. 2.1.2. Extended filtration The 5 L AnMBR underwent 3 months of acclimatization until the solids concentration was steady at around 2300 mg TSS/L before the experiment commenced. Hydraulic and solids retention times were fixed at 11.7 h and 50 days, respectively. Through feedback control of the permeate suction pump, a flux of 8.5 LMH was maintained, inclusive of 2 min of relaxation for every 8 min of suction. 3 N HCl or 3 N NaOH was automatically dosed to maintain the pH at 7.070.5. Inflow, membrane flux, and liquid level were regulated by a programmable logic controller (Renu Electronics, India) that executed 70 min of substrate feeding every 140 min. Temperature was separately controlled at 30–32 1C, using a heating tape and controller (TOHO Electronics, Japan). Biogas

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

Level

pH

Temperature

Produced biogas

Gas bag

101

Computer (datalogger)

Gas flowmeter Pressure gauge

Biogas recirculation

Permeate pump

Liquid flowmeter

Influent pump

Feed tank

Permeate tank

Gas pump

Fig. 1. Schematic of 5 L AnMBR. Thick and thin solid lines indicate liquid and gas flows, respectively. Dashed lines indicate data transfers.

production, permeate flowrate and TMP were measured by a thermal mass flowmeter (EW-32707-00, Cole Parmer, USA), a turbine flowmeter (S111, McMillan, USA) and a pressure transducer (EW-68075-02, Cole Parmer, USA), respectively. The entire experimental period was divided into three 30-day phases: a control period without enzyme augmentation (Phase I), and test periods with membrane-immobilized enzymes (Phase II) and dispersed enzymes (Phase III). In Phase III, purified hydrolases at a dose of 3.6 mL/g influent COD was supplemented daily, after sludge and permeate samples were taken. The dosing regimen was adopted to minimize process perturbation, while replenishing the bioreactor with fresh enzymes to sustain their catalytic activities. Sampling was generally conducted weekly throughout all three phases. Fouled membranes were replaced with virgin ones at the end of each phase, during which precaution was taken to minimize oxygen intrusion by sparging the bioreactor headspace with nitrogen gas. Cake layers were gently removed from the membranes using a spatula, and the remaining gel layers were detached ultrasonically in a 37 kHz ultrasonic bath (Fisher Scientific, USA) for 30 min at 25 1C. Both the cake and gel layers were reconstituted in 50 mL phosphate buffered saline (PBS; Crystal PBS buffer, Bioline, UK) by gentle vortexing. 2.2. Enzyme preparation and immobilization procedure Crude enzyme blend (BioCat Microbials, USA) containing proteases, lipases and amylases (21% w/v) was purified using a 100 mL stirred cell equipped with a 10 kDa membrane (Amicon PM 10, Millipore, USA) to remove low molecular weight additives. The cell was immersed in an ice bath and filtration proceeded under a nitrogen pressure of 200 kPa. Thereafter, the retained enzyme pellets were reconstituted in PBS to the original volume. Protease activity in the purified blend was 10 mmol/min/mL as determined by a commercial protease protocol (SSCASE01.001, Sigma Aldrich, USA), whereas amylase activity was 6.3 mg/min/mL, as determined from the 3,5-dinitrosalicylic acid method [32]. The enzyme immobilization procedure was adapted from Lozano et al. [33]. Virgin membranes were coated with gelatin by filtering a 1% w/v solution under 200 kPa for 1 h at 25 1C, followed by removing the excess using deionized water. This forms a pressure-compacted gelatin layer. Covalent crosslinking with glutaraldehyde (2% v/v) was conducted at 30 1C for 6 h under 150 rpm of shaker agitation (Multitron Standard, Infors HT, Switzerland). After washing off excess glutaraldehyde, the treated

membranes were soaked in the enzyme blend for 20 h at 6 1C. For the extended filtration experiment, 100 mg/L of acylase (A8376, Sigma-Aldrich, USA) was also added to minimize the biodegradation of immobilized enzymes [6]. Excess enzymes were washed off using PBS. The specific protease and amylase activities on the membrane surfaces were 59 mmol/min/m2 and 42 mg/min/m2, respectively. 2.3. Chemical analyses TSS was quantified according to Standard Method 2540D [34]. Dissolved organic carbon (DOC) was determined using the TOC ASI-V analyzer (Shimadzu, Japan). Soluble samples including SMP were obtained by centrifuging sludge suspensions at 3350g for 20 min at 4 1C, followed by filtering the supernatants through 0.45 mm syringe filters (Supor membrane, Pall, USA). Sludge-bound EPS was extracted using cation exchange resin (CER; Dowex Marathon C, Na þ form, Merck, Germany), by adapting the method described by Frølund et al. [35]. 50 mL of sludge suspension was centrifuged at 3350g for 20 min at 4 1C. The supernatant was decanted, followed by resuspending the pellets using PBS. This washing procedure was repeated twice. The washed suspension was transferred to a 50 mL glass bottle containing 80 g CER/g TSS, and chilled in an ice bath under magnetic stirring at 600 rpm for 1.5 h. The extracted EPS was recovered as the supernatant after centrifugation at 17,400g for 20 min. Protein concentration was measured using the microbicinchoninic acid protein assay kit (Thermo Scientific, USA), whereas carbohydrate concentration was determined by adapting the modified phenol–sulfuric acid method [36]. For the latter, absorbance was measured at 490 nm after reacting the acidified solution with 80% phenol for 30 min at 25 1C. 2.4. Instrumental analyses 2.4.1. Particle size distributions Particle size distributions (PSDs) of the sludge suspensions were obtained using a particle size analyzer (Mastersizer 2000, Malvern, England) with a detection range of 0.02–2000 mm. Each sample was stirred at 1350 rpm, and three repeated measures were used to construct a volume-weighted size distribution. Smaller particles from the gel layers were analyzed by Zetasizer (Nano-ZS, Malvern, England), which has an analytical range of 0.3 nm–10 mm. Measurements were repeated 15 times, and the

102

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

scattered light signals were processed and converted to a volumebased distribution.

2.4.2. Scanning electron microscopy Micrographs of membranes were taken using EVO 50 scanning electron microscope (SEM; Carl Zeiss International, Germany), typically with an extra high tension setting of 10 kV. For crosssectional measurements, membranes were first cryogenically fractured in liquid nitrogen. Samples were fixed using 2% (v/v) glutaraldehyde for 1.5 h, followed by washing thrice with 0.1 M cacodylate buffer. The samples were then dehydrated in succession using 50%, 70%, 85% and 95% ethanol for 10 min each. Before analysis, the dried membranes were mounted onto SEM stubs using carbon tapes, and sputter-coated with platinum for 20–30 s at 20 mV.

2.5. Data processing The initial TMP of each filtration run was subtracted from temporal TMP values to aid comparison between runs. All TMPs reported in the results section were standardized accordingly. Loess smoothing on extended filtration data was performed with local linear least squares regression and a smoothing parameter of 0.2. The latter was selected to provide adequate smoothing without excessively attenuating data features. Wherever possible, analytical results were expressed as mean 7standard deviation.

3. Results and discussion 3.1. Membrane properties and filtration characteristics

2.4.3. Contact angle measurements Membrane contact angles were estimated using the sessile drop method at  25 1C with an optical goniometer (OCA 15EC, Dataphysics, Germany). Samples were dried for one day in a desiccator before analyses. 10 mL of distilled water droplet was transferred onto the membrane surface using a microliter syringe. An image was captured 5 s after the droplet contacted the membrane. The elapsed time selected was a compromise between the time required for the droplet to stabilize and the continuous liquid drainage into the membrane. The baseline and droplet profiles were automatically detected by the accompanying software (SCA 20, Dataphysics, Germany), and manually readjusted if necessary. Static contact angle was then determined from the extracted profiles. The measurement was repeated eight times at different locations on the membrane sample, carefully avoiding previously wetted areas.

2.4.4. Capillary flow porometry Mean pore diameter and pore size distribution were determined using a capillary flow porometer (CFP 1500A, Porous Materials, USA), based on the bubble point method. The membranes were soaked in 1,1,2,3,3,3-hexafluoropropene (Galwick, Porous Materials, USA) for one day prior to analysis. After securing the precut membrane onto the holder, nitrogen pressure was progressively increased to displace the wetting liquid from the pores. Pore size was determined from the applied pressure in accordance with the Young–Laplace equation. Mean flow pore diameter was determined at the mean flow pressure, at which half of the nitrogen flow passes through larger pores [37].

Changes to the HVLP membrane following enzyme immobilization were examined to identify possible confounders on filterability. SEM images of the HVLP and the modified membranes are shown in Fig. 2(a) and (b), respectively. The HVLP membrane surface had a thick polymeric network that was rough and porous. At the top were distinct circular structures 10–20 μm in diameter. Within and between these structures were valleys which embedded channels 1–3 μm in diameter. Membrane pores lay within these channels. The modified membrane retained the general morphology of the HVLP. There were slight textural changes to the polymeric fibers making them appear marginally thicker or fuller. This did not substantially affect the flow channels and the architecture remained porous. The coat of immobilization mixture on the membrane appeared fairly uniform which expectedly resulted in an increase in the contact angle from 110 731 to 128 731. There was a marginal reduction in the mean flow pore diameter from 0.43 to 0.41 mm after enzyme immobilization. The most commonly detected pores (22–24% of all pores) remained in the 0.25–0.30 μm range (Fig. 3). The distribution of pores from 0.35 to 1.00 μm, encompassing  70% of all pores remained largely unchanged. However, there were marked reductions in the fraction of pores in the ranges of 0.25–0.30 μm (7%) and 0.30–0.35 μm (31%). These were proportionate with the emergence of pores between 0.20 and 0.25 μm, which constituted 8.4% of the total pores in the modified membrane. It was apparent that proteins from the immobilization mixture adsorbed onto and measurably constricted the smaller pores (0.25–0.35 mm) with greater ease, reducing them by at most 0.1 mm.

Fig. 2. SEM images of (a) virgin HVLP membrane and (b) HVLP membrane with immobilized enzymes.

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

25

Percentage of pores (%)

HVLP HVLP with immobilized enzymes 20

15

10

5

0. 10 0. - 0. 15 15 0. - 0. 20 20 0. - 0. 25 25 0. - 0. 30 30 0. - 0. 35 35 0. - 0. 40 40 0. - 0. 45 45 0. - 0. 50 50 0. - 0. 55 55 0. - 0. 60 60 0. - 0. 65 65 0. - 0. 70 70 0. - 0. 75 75 0. - 0. 80 80 0. - 0. 85 85 0. - 0. 90 90 0. - 0. 95 95 1. - 1. 00 00 -1 .0 5

0

Pore size (μm) Fig. 3. Pore size distributions of HVLP membranes with and without immobilized enzymes.

At the onset of filtering 2300 mg/L of sludge (at a flux of 8.5 LMH) when foulant attachment was minimal, the immediate filtration resistances of HVLP and the modified counterpart were both approximately 3.8  1012 m  1. This affirmed that the small morphological changes to the HVLP membrane did not substantially alter filtration characteristics. The ensuing divergence in membrane performances could therefore be attributed to physicochemical and enzymatic influences. 3.2. Batch filtration The initial assessment on the efficacy of hydrolases for fouling mitigation was carried out via batch filtration. As cake formation largely influence the long-term filterability in AnMBRs even under low flux [38,39], short-term supra-critical flux experiments were conducted to promote accelerated cake layer fouling. Although short- and long-term cake developments are likely to yield cakes with different properties [40], the propensity to form cakes is governed by similar physical and physicochemical factors. These experiments allow inference of the severity of cake fouling over longer time periods when operating at moderate flux. At a flux of 40 LMH, the control TMP increased slowly during the first 10 min before rapidly increasing thereafter (Fig. 4a). This signified plugging of large pores in the initial stage of filtration, followed by the continued plugging of smaller ones and the onset of cake formation [41]. Halving the flux reduced permeation drag and the filtrate volume per unit time, which extended the period of slow rising TMP to 20 min (Fig. 4b), but did not restrict the subsequent surge in TMP. Under the presence of 35 mL dispersed enzymes and a flux of 40 LMH, the increase in TMP during the entire experimental period was slow and steady (Fig. 4a), with an overall rate matching the first 10 min of the control TMP. The constant gradient indicated that large pores remained intact even after 50 min, and that the rise in TMP could be attributed to the development of an overlaying porous cake. The TMP response with 14 mL dispersed enzymes (20 LMH) was similar to that at the higher dose with a lower overall rate of TMP rise compared to its control (Fig. 4b). The slow ascent of both the control and the test TMPs during the first 20 min suggested that a significant amount of pores remained intact. Thereafter, pores were rapidly plugged in the control membrane and cake fouling commenced, as reflected by the rapid TMP increase. In the test reactor, not only did the more subdued TMP rise pointed toward a lower rate of cake formation, the relatively linear TMP profile especially after 30 min indicated the prevalence of intact

103

pores. As the cake layer did not result in excessive impedance to filterability, it was likely uncompacted and porous. These assertions were supported by SEM images of fouled membranes from the control and test AnMBRs. As shown in Fig. 5b, the fouled membrane from the control reactor was covered with a relatively dense cake possessing only a few distinct channel openings. The thin and elongated fissures could be artifacts resulting from sample drying, and should not be attached with too much importance. On the other hand, the fouled test membrane (Fig. 5a) retained the general morphology of the unused membrane (Fig. 2a), wherein the large valleys could still be discerned. There were signs that EPS had begun to create their own network, but channels through the cake were aplenty to the extent that the underlying membrane surface remained observable. One obvious difference between the TMP profiles with 35 and 14 mL dispersed enzymes was the greater concavity of the latter indicating a comparatively greater preponderance of pore plugging. Furthermore, despite operating at a lower flux of 20 LMH, the TMP after 44 min (Fig. 4b) was higher than that operating at 40 LMH (Fig. 4a), suggesting that enzyme dose could have a significant influence on membrane performance, with the impact greater at the later stage of a filtration cycle. The extent of fouling mitigation is more precisely described by ΔTMP, which is also depicted in Fig. 4 accompanied with best fit curves to aid visualization. ΔTMP of 0 corresponds to the separation point between the test and control TMPs. Apart from the interpretation of increasing percentage divergence of TMP t relative to TMP c , a rising ΔTMP also signifies a lower normalized rate of TMP increase of the test membrane (TMP 0t =TMP t ; “0 ” denotes the first derivative with respect to time) relative to the control ðTMP 0c =TMP c Þ. The period when this occurs closely corresponds to when fouling mechanisms substantially differ. When the two normalized rates are equalized, ΔTMP becomes constant and similar foulant deposition mechanisms contributing to cake development may be assumed. This occurred after at least 30 min at both doses. The eventual value of ΔTMP indicates the maximum possible improvement that can be anticipated under a specific set of conditions. This is dose dependent, and was around 60–70% for the higher dose and 50% for the lower one. The experiment on the immobilized enzyme and control pair had a shorter duration, owing to the substantial accumulation of decayed cell products in the batch-cultivated sludge causing rapid fouling. The control TMP rose sharply after 4 min and reached 40 kPa in 16 min (Fig. 4c). As with the dispersed hydrolases, the increase in TMP in the test reactor was moderated, rising to 21 kPa over the same period. Separation of the pressure profiles occurred at 2.5 min, and by 6 min, ΔTMP plateaued at around 50%. Hence, despite the adverse sludge characteristics, the enzymeimmobilized membrane performed better than the control membrane, and was comparable to the membranes contacting dispersed enzymes. The effectiveness of the two forms of enzymes is captured in the plot of retardation factor against TMP c (Fig. 4d). Points in the plot were derived from interpolated TMP data in Fig. 4(a)–(c). With dispersed hydrolases, the retardation factor increased monotonically with pressure. The initial rapid rise was attributed to the considerably slower fouling rate of the test membranes relative to the control membranes. As the normalized fouling rates converged at higher TMP, t t =t c increased at a diminishing rate, reflecting the persistent effect of dispersed enzymes. In contrast, the t t =t c of the enzyme-immobilized membrane first rose and then fell, highlighting the subtle uniqueness in the underlying TMP profile and therefore fouling development. It has to be emphasized that bulk sludge characteristics were altered with the application of dispersed hydrolases but not with immobilized ones. Since the same sludge was contacting the enzyme-immobilized membrane and its

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

100

40 Control (40 LMH) 35 mL dispersed enzymes ΔTMP

Control (20 LMH) 14 mL dispersed enzymes ΔTMP

80

60 20 40

TMP (kPa)

30

ΔTMP (%)

TMP (kPa)

30

100

40

10

80

60 20 40 10

20

0

0

10

20

30

40

20

0 50

0

0

10

20

Time (min)

40

0 50

2.5

100 Control (40 LMH) Immobilized enzymes ΔTMP

35 mL dispersed enzymes 14 mL dispersed enzymes Immobilized enzymes

80

20 40

2.0

tt /tc

60

ΔTMP (%)

30

30

Time (min)

40

TMP (kPa)

ΔTMP (%)

104

1.5 10 20

0

0

5

10

15

0 20

Time (min)

1.0

0

5

10

15

20

TMPc (kPa)

Fig. 4. TMP profiles of membranes contacting (a) 35 mL dispersed enzymes, (b) 14 mL dispersed enzymes, and (c) immobilized enzymes during batch filtration experiments, together with their respective controls. Retardation factor is plotted against TMPc in (d). Dashed lines are best-fit curves for ΔTMP and t t =t c .

control, foulants were deposited at about the same rate on both membranes, although a higher foulant deposition rate on the former was likely from a hydrophobicity perspective. Nonetheless, hydrolysis of cake macromolecules is a major efflux in the cake mass balance of enzyme-immobilized membranes, and this was manifested in the region leading to the t t =t c peak in Fig. 4(d). The subsequent decline could in part be attributed to the decreasing accessibility of immobilized hydrolases on newly deposited cake materials as the cake thickened. Fig. 5(c) shows the fouled enzyme-immobilized membrane. The cake had a layered texture comprising of a flaky top with numerous channel openings, and a somewhat more compact second layer. The two layers seemed to be interweaved at some locations and segregated at others. In areas where the layers were separated, the top layer could be prone to detachment when operating under back-pulsing or suction–relaxation filtration schemes. The fouled control membrane (Fig. 5d) had a very dense and compact cake with few channel openings, similar to the control membrane used in the dispersed enzyme experiment (Fig. 5b). 3.3. Extended filtration The efficacy of exogenous enzymes under short periods of supra-critical flux filtration was demonstrated in the batch experiments. However, their effectiveness under extended periods of continuous filtration required additional investigations. Fig. 6 presents the 30-day TMP time series for each phase. In the control

phase, the TMP increased almost linearly in the first 5 days at a rate of about 10 kPa/d, before rapidly leveling off to reach a relatively stable TMP of 44 kPa. In Phase II, the TMP ascended at a faster rate of 16 kPa/d in the first 2 days, reaching 31 kPa before tapering off. The higher relative rate of TMP increase stands in contrast to the lower relative rate of TMP rise encountered in the batch experiment. However, it is not possible to present a definitive explanation as any of the factors that differed between the batch and extended filtration experiments (gas superficial velocity, presence/absence of flux relaxation, net flux), could contribute to the discrepancy. Nevertheless, this should not detract from the other important observation that a lower pseudo-stable TMP was reached more quickly compared to Phase I, indicating the significant contribution of foulant hydrolysis toward lowering the cake resistance and a more rapid attainment of pseudo-equilibrium. To confer a lower resistance, it is necessary for the cake layer to be either thinner or more porous, or both, compared to that in Phase I. This would be in line with the microscopic observations in the batch experiment. The average rate of change in the ensuing 27 days was less than þ0.3 kPa/d. During this period, there were segments in which the TMP displayed higher than average increases followed by short and steep declines. For instance, from Day 8 to 13, the Phase II TMP rose from 36 to 41 kPa from the buildup of sludge cake. The subsequent drop to 37 kPa by Day 14 was speculated to be due to the shear-induced erosion of sludge cake weakened at the base as a result of EPS hydrolysis. The ascent–descent cycle also occurred

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

105

Fig. 5. SEM images of fouled membranes from batch filtration experiments. Images on the left are (a) membrane contacting 14 mL dispersed enzymes, and (c) enzymeimmobilized membrane. Their respective controls are displayed on the right.

25

25 SMP-protein EPS-protein

SMP (mg/L)

TMP (kPa)

40

30

20

10

0

Phase I (control) Phase II (immobilized enzymes) Phase III (dispersed enzymes) 0

5

10

15

20

25

SMP-carbohydrate EPS-carbohydrate

20

20

15

15

10

10

5

5

0

0

30

Time (day) Fig. 6. TMP profiles during Phase I (control period), Phase II (membranes with immobilized enzymes), and Phase III (dispersed enzymes) of extended filtration. Loess-smoothed lines illustrate underlying trends.

during Day 17–22 and Day 25–29, suggesting some persistence in the effects of immobilized enzymes. Unfortunately, these cycles did not translate into a significantly lower TMP compared to that in the control phase. The average difference in the TMPs after Day 10 was around 4 kPa, similar to the estimated error of the pseudosteady average TMP of each phase. Whereas the modified membranes displayed efficacy for less than 10 days, fouling alleviation by the dispersed hydrolases was apparent throughout the experimental period. The Phase III TMP increased at a much slower rate compared to the previous phases. It was below 10 kPa for the first 3 days and reached 22 kPa by Day 5, approximately 3 days longer than what Phase I filtration took to reach a similar level. Also noteworthy was the rapidity in which

EPS (mg/g TSS)

50

Phase I (control)

Phase II (immobilized enzymes)

Phase III (dispersed enzymes)

Fig. 7. Component concentrations of EPS and SMP in the various phases of the extended filtration experiment. Error bars represent standard deviations from 4 samples measured weekly.

the enzymes impacted filtration, affecting the TMP as soon as Day 1. The TMP transitioned to a gentler gradient from Day 5 to reach 29 kPa on Day 15. The average TMP during this period was 26 kPa, much lower than the corresponding Phase I average TMP of 44 kPa. It was evident that enzyme-induced alteration of the sludge had drastically modified its filterability and the associated development of the cake layer. Between Day 16 and Day 17, the TMP rose steeply to around 36 kPa and persisted for 3 days. A similar but shorter episode was apparent on Day 26. These increases in TMP were due to operating anomalies, which upon resolution, returned the system to normalcy. The absence of irreversible impairment demonstrated the resilience of the process. Despite these operating imperfections, the effective TMP was still nearly 30% lower than that of Phase I.

106

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

3.3.1. EPS and SMP concentrations Quantitative comparisons of bound EPS and SMP during extended filtration are illustrated in Fig. 7. EPS is plotted on a per g TSS basis to be in line with the conventional presentation of EPS derived from bulk sludge. As the reactor TSS was unchanged in Phases I and II, and declined by only 200 mg/L in Phase III, a volumetric EPS (in mg/L) would have largely the same relative composition amongst phases. During Phase I, EPS–protein and –carbohydrate concentrations were 1872 mg/g TSS and 3.470.3 mg/g TSS, and SMP–protein and – carbohydrate concentrations were 1573 mg/L and 3.070.5 mg/L, respectively. The component concentrations were similar during Phase II. On the other hand, after supplementing dispersed enzymes, EPS– protein decreased to 1273 mg/g TSS signifying the hydrolysis of the protein component of EPS into SMP. The simultaneously occurring biodegradation of proteinaceous SMP (including the slowly biodegradable exogenous enzymes [17]) limited the SMP–protein concentration to 1972 mg/L—a non-substantial increase compared to the preceding phases. EPS–carbohydrate appeared to be somewhat more resistant to hydrolysis as the concentration remained unchanged (3.271.0 mg/g TSS). EPS thus became less proteinaceous and more carbohydraceous in Phase III. SMP–carbohydrate registered an increase to 5.272.4 mg/L. The additional carbohydrates could be originated from the intracellular contents of 200 mg/L of lysed cells as inferred from the reduction in TSS. Although the amount was small, it was sufficient to contribute to the soluble contents especially if they were not readily biodegradable. With respect to the total soluble organic concentration, changes in Phase III were marginal (2272 mg DOC/L) compared to that in Phases I and II (1971 mg DOC/L). Therefore, any increase in contribution to fouling by the soluble organics during Phase III was likely small, and more than compensated by the reduction in bound EPS, primarily the proteinaceous component. At the end of Phase I, the cake density on the fouled membrane was 7.8 71.2 mg TSS/cm2. That of the enzyme-immobilized membrane at the end of Phase II was 4.8 70.8 mg TSS/cm2. Supplementation with dispersed enzymes (Phase III) led to the lowest cake buildup of 4.3 70.9 mg TSS/cm2. These alone could explain in large part the better membrane performance in Phase III. However, the similar cake densities in Phases II and III coupled with the better TMP profile of the latter (Fig. 6) suggested that the resistance of the fouled enzyme-immobilized membrane was dependent on other factors. An analysis of the composition of the cake and gel layers of fouled membranes is summarized in Fig. 8. Throughout the phases, the cake layer derived SMP–protein, EPS–protein and EPS–carbohydrate were similar. The components were in tight ranges of 19–21 mg/g TSS, 10– 12 mg/g TSS and 4.7–5.2 mg/g TSS, correspondingly. SMP–carbohydrate from the Phase II cake (1273 mg/g TSS) was not remarkably

30 SMP-protein (gel) SMP-protein (cake) EPS-protein (cake)

SMP-carbohydrate (gel) SMP-carbohydrate (cake) EPS-carbohydrate (cake)

25

0.06 20 0.04

15 10

0.02 5 0.00

Cake layer SMP, EPS (mg/g TSS)

Gel layer SMP (mg/cm2)

0.08

0 Phase I (control)

Phase II Phase III (immobilized enzymes) (dispersed enzymes)

Fig. 8. Component concentrations of EPS and SMP in cake layers, and SMP in gel layers. Error bars represent standard deviations obtained from sampling three random locations on fouled membranes.

higher compared to the other phases (Phase I: 971 mg/g TSS; Phase III: 973 mg/mg TSS). As such, the cake composition was very similar throughout the three phases. Unlike the cake, the composition of the gel layer displayed greater disparities. Both the Phase III SMP–protein and –carbohydrate were higher than in Phase I. The Phase II SMP–carbohydrate was unremarkably low and similar to Phase I, and not significantly different from Phase III. SMP–protein in Phase II was significantly higher (0.04670.006 mg/cm2) compared to Phase I (0.006 70.003 mg/cm2) and Phase III (0.015 7 0.008 mg/cm2). The very high SMP–protein of the Phase II gel layer was therefore the most probable culprit in imparting additional resistance to the fouled enzyme-immobilized membrane and can be appreciated from a heuristic approximation of the gel resistance. Several simplifying assumptions were made including (1) linear dependence between cake resistance and cake density, (2) linear dependence between gel resistance and gel areal protein concentration (carbohydrate disregarded), (3) similar resistanceimparting cake characteristics in all phases, and (4) similar resistance-imparting gel properties in all phases. These enabled the following to be written: RT;i Rm;i ¼ α½cake densityi þ β ½gel layer proteini :

ð2Þ

The left hand side of Eq. (2) is the difference between the total (RT;i ) and the membrane resistance (Rm;i ), both of which were known from the experiments. The terms on the right hand side represent the cake (Rc;i ) and gel (Rg;i ) resistances, with α denoting the specific cake resistance and β denoting the specific gel resistance. Subscript i represents the phase. The two unknowns, α and β, were solved using information from Phase I and III to give 2.4  108 m/mg TSS and 2.8  1010 m/mg, respectively. Phase II gel resistance was estimated from Rg;II ¼ ðRT;II  Rm;II Þ  α½cake densityII ; 12

ð3Þ

1

yielding a value of 7.0  10 m , which was 60% of the cake resistance. The high percentage stands in contrast to the 9% for Phase I (1.7  1012 m  1 gel resistance). Phase III too had a fairly high percentage (40%; 4.2  1012 m  1 gel resistance), that was accompanied with a relatively low cake resistance (1.0  1013 m  1). Therefore, even though the enzyme-immobilized membrane had a low cake density, its performance was hampered by the high gel resistance. The possibility that the gel extracting ultrasonication procedure (Section 2.1.2) removed a portion of the gelatin–enzyme coating should be considered. This would have resulted in an overestimation of the gel protein concentration and was the reason for approximating Rg;II from the cake density. A hypothetical uncontaminated gel layer protein concentration was estimated from Rg;II =β yielding 0.025 mg/cm2—a value that is still high at 1.7 times that of the Phase III concentration, and 4.0 times the Phase I concentration. Notwithstanding the procedural related uncertainty, the elevated protein concentration is significant and must be related to the more hydrophobic surface of the modified membrane. It is postulated that resistance was transferred from the cake to the gel as hydrolysis at the base of the cake resulted in hydrolysis products attaching strongly to the gelatin. 3.3.2. Particle size distributions Throughout the three phases, the floc sizes of the bulk sludge were predominantly in the 10–500 mm range (493% of all particles) with major peaks at around 50 mm and shoulders ( 15% of major peak height) at around 450 mm. Volumeweighted mean diameters ranged between 89 and 107 mm. Unlike in a previous work [17], where net floc agglomeration was observed following a sustained period of dispersed enzyme augmentation, a significant change in the particle size was not as evident in this study.

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

(by volume) in Phases I and III, respectively, were of sizes smaller than 0.25 μm, and could potentially enter any membrane pore. In addition, the size of the largest membrane pore (0.95–1.0 μm) coincided with the upper end of the range of submicron gel particles. These suggest that the majority of the ultrasonically dislodged materials had resided within the membrane pore network. Despite the weaker correlation between the Phase II PSD and the pore size distribution of the modified membrane, the dislodged particles possessed much of the size characteristics of Phase I particles. Every Phase II particle smaller than 2 μm appeared to be twice the size of Phase I particles; thereby, retaining Phase I’s relative volumetric contribution of each size class. There was also no indication of the emergence of a new group of particles with a distinct size distribution. As such, it was highly likely that foulants in the Phase II gel were similar to those in Phase I and had resided within the membrane pore network. Owing to the large surface-area-to-volume ratio, submicron particles would have had ample contact with the immobilization coating and engaged in considerable attractive interactions especially that related to the hydrophobic effect. If the total adhesion force was strong and gelatin in the vicinity had insufficient crosslinking (which may be possible over a small localized area), then during sonication, a proportionate amount of gelatin could be dislodged together with the foulant particles. Conversely, large 5 μm particles residing exterior of the membrane pore network would not have had much contact with gelatin due to their small surface-area-to-volume ratio. Coupled with the more extensive crosslinking in the gelatin over a large area, these particles were dislodged unaccompanied and without a change in size.

5

Volume (%)

4

Phase I (control) Phase II (immobilized enzymes) Phase III (dispersed enzymes)

3

2

1

0 0.1

1

10

100

1000

Particle size (μm)

18 16 14

Phase I (control) Phase II (immobilized enzymes) Phase III (dispersed enzymes)

Volume (%)

12 10 8 6 4 2 0 0.01

0.1

1

107

10

Particle size (μm) Fig. 9. PSDs of (a) suspended cake layers, and (b) sonicated gel layers from fouled membranes after 30 days of filtration.

Particle size differences between the phases were more apparent in the cake sludge. Fig. 9a depicts the PSDs of the suspended cake layers. The major peak of the control phase PSD was at 50 μm with a pronounced secondary peak at 480 μm. Shoulders of the Phases II and III PSDs were at around 500 μm. Both the major peaks of the Phase II and Phase III PSDs (centered at 90 and 70 μm, respectively) together with portions of the PSDs from 10 to 300 μm appeared to be right-shifted from the Phase I PSD, resulting in a substantial increase in the proportion of particles in the 80–400 μm range, and a reduction in the o60 μm range. Volume-weighted mean diameters were approximately 160 mm (Phase I), 220 mm (Phase II) and 170 mm (Phase III). Although enhanced enzymatic hydrolysis should reduce the size of macromolecules, it apparently does not translate into a size reduction of deposited flocs. It is conjectured that as enzymes act on the outer and more accessible region of flocs, the ensuing structural changes enable deposited and depositing flocs to coalesce more easily, either via more favorable steric or physical–chemical factors, or both. The PSD of the control phase gel layer had a major peak at 0.39 μm, and two secondary peaks at 0.07 and 5.6 μm (Fig. 8b). The PSD of the Phase II gel appeared to be right-shifted from the Phase I PSD for sizes less than 1.8 μm with the major peak at 0.8 μm. The larger-sized particles remained unchanged in size (  5 μm) but became slightly more abundant. In Phase III, only one major peak at 0.38 μm was present. The volume-weighted mean particle diameter of the gel layers of Phases I, II and III were 0.45, 0.90 and 0.36 μm, respectively. There were close correspondences between the PSDs of Phases I and III, and the HVLP membrane pore size distribution (Fig. 3). About 38% and 26% of gel particles

3.4. Operational implications and other remarks It has to be reiterated that the intended mode of action of dispersed and immobilized hydrolases are different. Dispersed enzymes act primarily on bulk sludge to alter materials prior to deposition, while the immobilized ones target deposited materials. The former requires enzymes to be regularly replenished to maintain an effective level of activity, and hence the cumulative amount used would always increase. To minimize dosage, it is necessary to contemplate the level of TMP reduction desired, since there may be a dose-dependent relationship (Section 3.2). The dosage may also be adjusted depending on the resilience of the microbial community in response to the pulse disturbance [42]. For instance, if it takes 5 days for the EPS–protein concentration to return to the pre-disturbance value, daily dosing might not be necessary. Implementation of such fouling control strategy must also consider the impact on the biological performance of AnMBR. Under the same operational conditions stated in Section 2.1.2, an earlier study [17] showed that solids accumulation was regulated when treating sewage containing organic particulates and permeate quality was consistently good. However, issues such as fluctuations in biogas production and enzyme biodegradation require further resolution. These can be addressed via a modification of the dosing regimen (e.g. smaller and more frequent doses) which may or may not be compatible with the optimal for fouling control, and thus may require a compromise. The aforementioned complexities are avoided in the enzymeimmobilized membrane. Although the membrane, with its fixed enzyme load, is not expected to remain active indefinitely, the effect on cake density may exhibit greater persistence (Section 3.3.1). The larger issue is with respect to the gel resistance arisen from the close proximity between the membrane and the hydrolysis sites. This may be alleviated with a less hydrophobic immobilization coating, but may not be eliminated without a path or process that enables the transport of hydrolysis products away from the

108

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

membrane. Finally, it should pointed out that the overall efficacy of the enzyme-immobilized membrane hinges on several factors such as the enzyme load (which governs the hydrolysis rate), the balance between the foulant deposition rate and the hydrolysis rate, the extent of flux relaxation (which promotes back transport of hydrolysis products particularly prior to the formation of a significant cake), the composition of the immobilization mixture, etc., all of which can be manipulated to certain extents. The performance of the enzyme-immobilized membrane reported in this work is specific to the conditions employed and may thus be improved with optimization.

4. Conclusions This study examined the efficacies of exogenous hydrolases in enhancing membrane performances in AnMBRs. Dispersed enzymes reduced the protein content of bound EPS in the bulk suspension, and restricted the membrane cake buildup. Higher SMP concentration in the gel layer relative to the control phase could be the inadvertent consequence of the low cake density as the membrane surface and pore network became more accessible to colloidal materials from the bulk solution. Although this led to a somewhat higher gel resistance, the greatly reduced cake resistance resulted in a nearly 30% reduction in the total resistance at pseudo-steady state. Batch filtration experiments conducted at very high flux also demonstrated the development of a porous and permeable cake. Moreover, a monotonically increasing retardation factor suggested that the superior performance of dispersed enzymes could be expected over a wide range of TMP. The localized influence of immobilized enzymes also brought about a low cake density and cake resistance comparable to that achieved with dispersed enzymes. However, the filtration performance of the enzyme-immobilized membrane was not substantially improved over the control membrane owing to an elevated gel resistance. Hydrolysis at the base of the cake apparently controlled the cake density, and could also have contributed proteinaceous hydrolysis products with strong affinities to the hydrophobic immobilization layer.

Acknowledgments This research was funded by the Singapore Membrane Technology Center, part of the Nanyang Environment and Water Research Institute. References [1] P.L. McCarty, J. Bae, J. Kim, Domestic wastewater treatment as a net energy producer—can this be achieved? Environ. Sci. Technol. 45 (2011) 7100–7106. [2] H. Lin, J. Chen, F. Wang, L. Ding, H. Hong, Feasibility evaluation of submerged anaerobic membrane bioreactor for municipal secondary wastewater treatment, Desalination 280 (2011) 120–126. [3] J. Kim, K. Kim, H. Ye, E. Lee, C. Shin, P.L. McCarty, J. Bae, Anaerobic fluidized bed membrane bioreactor for wastewater treatment, Environ. Sci. Technol. 45 (2010) 576–581. [4] A.Y. Hu, D.C. Stuckey, Activated carbon addition to a submerged anaerobic membrane bioreactor: effect on performance, transmembrane pressure, and flux, J. Environ. Eng. 133 (2007) 73–80. [5] I. Mishima, J. Nakajima, Control of membrane fouling in membrane bioreactor process by coagulant addition, Water Sci. Technol. 59 (2009) 1255. [6] K.-M. Yeon, W.-S. Cheong, H.-S. Oh, W.-N. Lee, B.-K. Hwang, C.-H. Lee, H. Beyenal, Z. Lewandowski, Quorum sensing: a new biofouling control paradigm in a membrane bioreactor for advanced wastewater treatment, Environ. Sci. Technol. 43 (2008) 380–385. [7] J. Wang, J. Guan, S. Santiwong, T.D. Waite, Characterization of floc size and structure under different monomer and polymer coagulants on microfiltration membrane fouling, J. Membr. Sci. 321 (2008) 132–138. [8] P. Le-Clech, V. Chen, T.A. Fane, Fouling in membrane bioreactors used in wastewater treatment, J. Membr. Sci. 284 (2006) 17–53.

[9] F. Meng, S.-R. Chae, A. Drews, M. Kraume, H.-S. Shin, F. Yang, Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material, Water Res. 43 (2009) 1489–1512. [10] S. Tsuneda, H. Aikawa, H. Hayashi, A. Yuasa, A. Hirata, Extracellular polymeric substances responsible for bacterial adhesion onto solid surface, FEMS Microbiol. Lett. 223 (2003) 287–292. [11] Y. Liu, H.H. Fang, Influences of extracellular polymeric substances (EPS) on flocculation, settling, and dewatering of activated sludge, Crit. Rev. Environ. Sci. Technol. 33 (2003) 237–273. [12] C.S. Laspidou, B.E. Rittmann, A unified theory for extracellular polymeric substances, soluble microbial products, and active and inert biomass, Water Res. 36 (2002) 2711–2720. [13] J. Wingender, T.R. Neu, H.-C. Flemming, Microbial Extracellular Polymeric Substances: Characterization, Structure and Function, Springer-Verlag, Berlin, 1999. [14] V.L. Puspitasari, M. Rattier, P. Le-Clech, V. Chen, Performances of protease and amylase cleaning for microporous membranes used in wastewater applications, Desalin. Water Treat 13 (2010) 441–449. [15] A. Maartens, E. Jacobs, P. Swart, UF of pulp and paper effluent: membrane fouling-prevention and cleaning, J. Membr. Sci. 209 (2002) 81–92. [16] F.I. Turkdogan-Aydinol, K. Yetilmezsoy, S. Comez, Effect of extracellular enzyme activity on digestion performance of mesophilic UASB reactor treating high-strength municipal wastewater, Bioprocess Biosyst. Eng. 34 (2011) 389–401. [17] C.W. Teo, P.C.Y. Wong, Enzyme augmentation of an anaerobic membrane bioreactor treating sewage containing organic particulates, Water Res. 48 (2014) 335–344. [18] E.V. Münch, P.C. Pollard, Measuring bacterial biomass–COD in wastewater containing particulate matter, Water Res. 31 (1997) 2550–2556. [19] A. Ayol, A. Filibeli, E. Kuzyaka, D. Sir, Biological sludge disintegration: fate and effects of hydrolytic enzymes for enhanced dewaterability, in: Proceedings of the IWA Specialist Conference on Moving Forward Wastewater Biosolids Sustainability: Technical, Managerial, and Public Synergy, 2007, pp. 527–534. [20] H.-Y. Pei, W.-R. Hu, Q.-H. Liu, Effect of protease and cellulase on the characteristic of activated sludge, J. Hazard. Mater. 178 (2010) 397–403. [21] I.P. Molobela, T.E. Cloete, M. Beukes, Protease and amylase enzymes for biofilm removal and degradation of extracellular polymeric substances (EPS) produced by Pseudomonas fluorescens bacteria, Afr. J. Microbiol. Res. 4 (2010) 1515–1524. [22] P. Hodgson, G. Leslie, A. Fane, R. Schneider, C. Fell, K. Marshall, Cake resistance and solute rejection in bacterial microfiltration: the role of the extracellular matrix, J. Membr. Sci. 79 (1993) 35–53. [23] J. Howell, O. Velicangil, Theoretical considerations of membrane fouling and its treatment with immobilized enzymes for protein ultrafiltration, J. Appl. Polym. Sci. 27 (1982) 21–32. [24] Y.-H.P. Zhang, S. Myung, C. You, Z. Zhu, J.A. Rollin, Toward low-cost biomanufacturing through in vitro synthetic biology: bottom-up design, J. Mater. Chem. 21 (2011) 18877–18886. [25] Y.-H.P. Zhang, Simpler is better: high-yield and potential low-cost biofuels production through cell-free synthetic pathway biotransformation (SyPaB), ACS Catal. 1 (2011) 998–1009. [26] T. Satyanarayana, B.N. Johri, A. Prakash, Microorganisms in Sustainable Agriculture and Biotechnology, Springer, 2012. [27] P.S. Nigam, A. Pandey, Biotechnology for Agro-Industrial Residues Utilisation: Utilisation of Agro-residues, Springer, 2009. [28] R. Tyagi, R.Y. Surampalli, S. Yan, T.C. Zhang, C. Kao, B. Lohani, Sustainable Sludge Management: Production of Value Added Products, ASCE Publications, 2009. [29] M.A. Tarr, Chemical Degradation Methods for Wastes and Pollutants: Environmental and Industrial Applications, CRC Press, 2003. [30] M. Kim, C.Y. Gomec, Y. Ahn, R. Speece, Hydrolysis and acidogenesis of particulate organic material in mesophilic and thermophilic anaerobic digestion, Environ. Technol. 24 (2003) 1183–1190. [31] P. Le Clech, B. Jefferson, I.S. Chang, S.J. Judd, Critical flux determination by the flux-step method in a submerged membrane bioreactor, J. Membr. Sci. 227 (2003) 81–93. [32] P. Bernfeld, Amylases: α and β methods, Enzymology 1 (1955) 149–158. ́ T. De Diego, D. Gómez, D. Paolucci-Jeanjean, [33] P. Lozano, A. Pérez-Marın, M. Belleville, G. Rios, J. Iborra, Active membranes coated with immobilized Candida antarctica lipase B: preparation and application for continuous butyl butyrate synthesis in organic media, J. Membr. Sci. 201 (2002) 55–64. [34] APHA, Standard Methods for the Examination of Water and Wastewater, 20th ed., American Public Health Association, Washington, DC, USA, 1998. [35] B. Frølund, R. Palmgren, K. Keiding, P.H. Nielsen, Extraction of extracellular polymers from activated sludge using a cation exchange resin, Water Res. 30 (1996) 1749–1758. [36] P. Rao, T.N. Pattabiraman, Reevaluation of the phenol–sulfuric acid reaction for the estimation of hexoses and pentoses, Anal. Biochem. 181 (1989) 18–22. [37] D. Li, M.W. Frey, Y.L. Joo, Characterization of nanofibrous membranes with capillary flow porometry, J. Membr. Sci. 286 (2006) 104–114. [38] D. Jeison, W. Van Betuw, J. van Lier, Feasibility of anaerobic membrane bioreactors for the treatment of wastewaters with particulate organic matter, Sep. Sci. Technol. 43 (2008) 3417–3431. [39] Y. Liu, H. Liu, L. Cui, K. Zhang, The ratio of food-to-microorganism (F/M) on membrane fouling of anaerobic membrane bioreactors treating low-strength wastewater, Desalination 297 (2012) 97–103.

P.C.Y. Wong et al. / Journal of Membrane Science 491 (2015) 99–109

[40] S.-H. Ou, S.-J. You, Y.-C. Lee, Extracellular polymeric substance characteristics and fouling formation mechanisms in submerged membrane bioreactors, Desalin. Water Treat. 18 (2010) 175–181. [41] H. Lin, K. Xie, B. Mahendran, D. Bagley, K. Leung, S. Liss, B. Liao, Factors affecting sludge cake formation in a submerged anaerobic membrane bioreactor, J. Membr. Sci. 361 (2010) 126–134.

109

[42] A. Shade, H. Peter, S.D. Allison, D.L. Baho, M. Berga, H. Bürgmann, D.H. Huber, S. Langenheder, J.T. Lennon, J.B.H. Martiny, K.L. Matulich, T.M. Schmidt, J. Handelsman, Fundamentals of microbial community resistance and resilience, Front. Microbiol. 3 (2012) 417.