A novel In-situ Enzymatic Cleaning Method for Reducing Membrane Fouling in Membrane Bioreactors (MBRs)

A novel in-situ enzymatic cleaning method was developed for fouling control in membrane bioreactors (MBRs). It is achieved by bringing the required enzymes near the membrane surface by pulling the enzymes to a magnetic membrane (MM) surface by means of magnetic forces, exactly where the cleaning is required. To achieve this, the enzyme was coupled to a magnetic nanoparticle (MNP) and the membrane it self was loaded with MNP. The magnetic activity was turned by means of an external permanent magnet. The effectiveness of concept was tested in a submerged membrane filtration using the model enzyme-substrate of Bacillus subitilis xylanase-arabinoxylan. The MM had almost similar properties compared to the unloaded ones, except for its well distributed MNPs. The enzyme was stable during coupling conditions and

A novel in-situ enzymatic cleaning method was developed for fouling control in membrane bioreactors (MBRs). It is achieved by bringing the required enzymes near the membrane surface by pulling the enzymes to a magnetic membrane (MM) surface by means of magnetic forces, exactly where the cleaning is required. To achieve this, the enzyme was coupled to a magnetic nanoparticle (MNP) and the membrane it self was loaded with MNP. The magnetic activity was turned by means of an external permanent magnet. The effectiveness of concept was tested in a submerged membrane filtration using the model enzymesubstrate of Bacillus subitilis xylanase-arabinoxylan. The MM had almost similar properties compared to the unloaded ones, except for its well distributed MNPs. The enzyme was stable during coupling conditions and the presence of coupling could be detected using a high-performance anionexchange chromatography (HPAEC) analysis and Fourier transform infrared spectroscopy (FTIR). The system facilitated an in-situ enzymatic cleaning and could be effectively applied for control fouling in membrane bioreactors (MBRs).

INTRODUCTION
Membrane bioreactors (MBRs) can be broadly defined as systems that integrate a conventional activated sludge system with a membrane filtration. (Judd, 2003, Visvanathan et al., 2000. MBR offers a better effluent quality and a more robust technology. MBRs have thus become widely accepted as an advanced wastewater treatment option. However, its widespread application is still restricted by a membrane fouling phenomenon, which lowers the applied flux, reduces the sustainability of operation, increases the frequency of maintenance and intensive cleaning, reduces the membrane lifetime, and eventually increases the investment and operational cost. The traditional approaches for fouling control are mainly based on the physico-chemical principles, such as membrane modification, optimization of operational conditions, coagulant addition, cyclic filtration, etc, (Le-Clech et al., 2006, Yeon et al., 2009a, Meng et al., 2009) which may not be effective and less energy efficient. Tackling the fouling remains one of the major challenges to expand its application and significantly improve its general acceptance by industries. Figure 1 shows the mechanism and the occurence of TMP jump. In general, there are three stages of fouling (Zhang et al., 2006), which are initial short-term rapid rise in TMP (stage 1 in Figure 1), a long-term weak rise in TMP (stage 2), followed by a TMP jump (stage 3). The occurrence of the TMP jump has been explained by several mechanisms: an increment of local flux exceeding the critical flux (CF) (Cho & Fane, 2002), a sudden change of biofilm or cake structure (Zhang et al., 2006), a sudden increase in the concentration of EPS at the bottom of the cake layer (Hwang et al., 2008), or the presence of N-acyl home serine lactone (AHL) as an association to biofilm formation (Yeon et al., 2009a). If this is critical time when the TMP starts to jump accursed, an interruption of the filtration will occur. The overall goal of fouling control is thus to retard the occurrence of the TMP jump. (Meng et al., 2009).
Activated sludge behaves as a complex and dynamic system: the TMP jump mechanisms are diverse and may in fact all occur simultaneously. The cake layer formation is initiated from the attachment of biopolymers, which starts to occur in the first stage and evolves and propagates in the second stage (See Figure 1). At the end of the second stage, when the local flux approaches the CF, the cake layer structure changes and an increase of EPS occur. Therefore, stage 2 plays a very important role in the occurrence of TMP Jump. More detailed investigation and properly handling of this stage can be the key step for the fouling control.
It has been widely accepted that the major fouling contributor is a cake layer that forms on the membrane surfaces (Chu & Li, 2006;Wang et al., 2007;Meng & Yang, 2007;Ramesh et al., 2007). This cake is formed from deposited biopolymers, which further allows easier and accelerated bacterial adhesions and holds the microbial flock on the membrane surface. A part of this cake reversibly detaches due to backtransport but the remaining cells multiply and excrete microbial products, which eventually clog the pores and form a strongly attached and irremovable fouling layer (Herzberg et al., 2009). The deposited biopolymers consist of dissolved and colloidal extra-cellular polymeric substances (EPS) and soluble microbial products (SMP) (Wang et al., 2008). EPS and SMP are produced by the microorganisms; they consist of a whole group of macromolecules, such as polysaccharides, proteins, nucleic acids, (phospho-)lipids, and other polymeric compounds (Le-Clech et al., 2006). These compounds facilitate bacterial attachment to membrane surface and act as a shelter to protect the bacteria in the biofilm-EPS matrix. EPS/SMP can be hydrolyzed by some specific enzymes, indicating an advanced way to control membrane fouling. Extracellular proteins and polysaccharides (as the main building blocks of EPS) can be degraded by proteolitic enzymes as has been successfully used for biofilm detachment (Loiselle & Anderson, 2003). In fact, enzymatic cleaning has long been applied for an intensive cleaning of fouled membranes (Maartens et al., 1996). Recent studies by Yeon et al. showed the effectiveness of quorum quenching enzyme (acylase) coupled onto magnetic nanoparticles for preventing the formation of biofilm in MBR (Yeon et al., 2009).
These results enable us to hypothesize that, in principle, membrane fouling originating from cake layer formation could also be alleviated through interrupting the initiation stage by enzymatic disruption of biopolymers. To achieve this goal, the concept of a novel in-situ enzymatic cleaning method called "magnetic enzymemagnetic membrane" (MEMM) is introduced and evaluated in this study.
Trans membrane pressure In the MEMM system, the enzymes are coupled to a super paramagnetic nanoparticle to form the magnetic enzyme (ME), while the membrane is loaded with super paramagnetic nanoparticles (magnetic membrane, MM). When an external magnetic field is applied to the superparamagnetic nanoparticles, they tend to align along the magnetic field, leading to a net magnetization. In that way, applying a magnetic field to the membrane will facilitate the attraction of the ME onto the MM surface by simple magnetic attraction. The MEs are then expected to break down the foulants locally, precisely where it is needed (not in the bulk sludge solution but only on the membrane surface). Importantly, the ME can be easily released from the membrane surface by switching-off the magnet in a later stage when for instance an additional chemical cleaning of MM is required. This will prevent the destruction of the MEs by the aggressive chemical cleaning agents. After the intensive membrane cleaning, the MEs can be reattracted to the MMs by simply reactivating the magnetic attraction. Since the MEMM operates very locally, the necessary amount of enzyme is expected to be much lower. As the cleaning process occurred continuously throughout the operation, this could be a more direct approach for fouling control in MBRs. This approach is non-toxic, environmentally benign and highly specific compared to aggressive chemicals that may also oxidize the membrane (Puspitasari & Liany, 2009) to shorten the membrane life span and damage the environment.
To prove the effectiveness of this novel MEMM concept, a model arabinoxylane substrate was used together with endo-β-1,4-D-xylanase. To achieve this objective, the research was conducted by the following steps: (1) Preparation of magnetic nanoparticles (MNP) that exhibit super paramagnetic properties, (2) Preparation and characterization of MM and evaluation of their magnetic properties, (3) Characterization of the fresh enzyme, (4) Preparation of ME and optimization of the immobilization process, (5) Performance evaluation or system tests.

Membrane preparation and module potting
The flat sheet membranes were prepared from polysulfone (PSF) (BASF-Ultrason)/N-Methyl-2-pyrrolidone (NMP) (ACROS) solutions (12, 14 and 16%) by phase inversion (Vankelecom et al., 2004). The polymer solutions were cast with a 250 µm wet thickness of casting knife on a polypropylene support (Novatexx 2471, kindly supplied by Freudenberg, Germany) and casting speed of 2.25 cm/s. The casting was performed in a semi-closed hood at ambient temperature (18-20°C) and a relative humidity of 20-30%. The casted polymer solution was immediately immersed into a coagulation bath containing demineralized water as non-solvent. Afterwards, the membrane was immediately stored in tap water prior to be used. To prepare MM, a series of 12% of PSF/NMP solutions were added with 10%, 20% and 30% (%w) of MNP. The relatively high MNP concentrations were chosen to obtain sufficient magnetic attraction force. To prevent MNP aggregation, the solvent was divided into two parts: (i) to dissolve the polymer (polymer solution) and (ii) to dissolve the MNP (MNP solution). The MNP solution was placed into a sonification device for four hours to break the aggregation. Subsequently, the polymer and MNP solutions were mixed and immediately stirred by using a mechanical stirrer and sonificated just before casting. Prior to use in system test, the MM was potted according to our previous reference (Bilad et al., 2010).
To synthesize approximately 100 mg of MNP, 37.5 mL of ethylene glycol and 37.5 mL of octylamine were first mixed in a reactor placed in an oil bath at 140°C and added with a mixture of 2.4 g of iron chloride, 10 mL of octylamine and 4 mL of distilled water (mixed at 50°C) by using a pipette. When the solution started to evaporate, the temperature was increased to 175°C, and until then the reflux was captured and kept for 24 hours. The color of the solution changed from brown to black, indicating the formation of MNP. To separate MNP from the solution, the solution was poured into a beaker glass and was added with acetone (act as non solvent for MNP). By placing a magnet at the outside-bottom of the beaker glass, the MNP settled at the bottom. The MNP were separated from the liquid phase by decantation. This procedure was repeated several times followed by vacuum dried at room temperature to obtain dry MNP powder. The hydrophilization of the MNP was done by addition of 1 mL of xylene (3-(bis(2-hydroxyethyl)amino)prophyltriethoxysilane) and few drops of acetic acid in MNP/acetone solution (±100 mL). This solution was sonicated for one hour to break the MNP aggregate following by the addition of distilled water until two phases solution were formed (the hydrophilic MNP moved to the water phase), which eventually could be separated by decantation. The MNP were recovered by vacuum drying. The MNP have a typical size of approximately 15±5 nm with difference ratios of acid to hydroxyl groups (0.001, 0.01, 0.05 and 0.1) obtained from further treatments.

Enzyme, substrate, and enzyme characterization
The model enzyme used in this study was Endo-β-1,4-D-xylanases (BsXynA) (EC 3.2.1.8). The further characterics of the enzyme were given by references (Vandermarliere et al., 2008;Pollet, 2010). The BsXynA was prepared in Laboratory of Food Chemistry and Biochemistry and Leuven Food Science and Nutrition Research Centre (LFoRCe), Katholieke Universiteit Leuven. The substrate of BsXynA used in this study was arabinoxylan.
(1) Determination of the pH optimum The pH optimum of BsXynA was determined by measuring enzyme activities at different pH levels. The xylazyme tablet (substrate) was placed into the tubes containing different buffer solutions: sodium acetate (NaOAc) buffers at pH 3, 4, and 5; sodium phosphate (Na 2 HPO 4 )/sodium biphosphate (NaH 2 PO 4 ) buffers at pH 6, 7 and 8 (VWR International, Belgium). The enzyme activity was measured by performing a colorimetric xylazyme activity test on the different test tubes (Section 0) contaning solution with different pH. To check the pH stability, the BsXynA was stored in different pH: 3, 4, and 5 in NaOAc buffer and pH 6, 7, and 8 in Na 2 HPO 4 /NaH 2 PO 4 buffer for 2 hours at room temperature prior to be used. Subsequently, the pH-values were adjusted to optimal pH by diluting it in MOPS buffer (pH = 6.5; 20 mM) for relative activity tests.
(2) Determination of the temperature optimum and stability The temperature optimum of BsXynA was determined by measuring the activities at its optimum pH. The reaction tubes were placed in water baths at temperatures of 20, 30, 40, 50, 60, and 70°C during the incubation. The activity of the enzyme was measured using Xyalazyme method. For temperature stability, the BsXynA was stored in different temperatures of 20, 30, 40, 50, 60, and 70°C for 40 minutes prior to be used. Subsequently, the activity of BsXynA was measured at temperature of 40°C and pH of 6.5 in a MOPS buffer (20 mM) for relative activity tests.

Coupling of BsXynA to MNP and optimization of coupling procedure
The MNP was coupled to the enzyme by covalently bind the carboxylic acid (-COOH) groups on the MNP to the amine (-NH 2 ) groups on the enzyme using N-(3dimethylaminopropyl)-N'-ethylcarbodiimide (EDC) (Sigma-Aldrich) and Nhydroxysuccinimide (NHS) (Sigma-Aldrich) cross-linking agents. The coupling procedure consists of two steps: activation (step 1) and reaction (step 2). To obtain the maximum coupled enzyme yield, the coupling was performed in different parameters: the ratio of carboxylic acid to amine groups of: 0.1, 1, 5, 10 and 100%; ratio of enzyme to MNP (%wt/%wt) of 10, 1 and 0.1; reaction time of 2 and 24 hours; reaction buffers; and the storage solutions. All experiments were preformed in triplicate.

Experimental set-up and system test
To prove the concept of the in-situ cleaning technique, the system tests were carried out by using prepared MM and ME in lab-scale set-up. The tests were consisted of three parts: (1) Ability of magnetic attraction of ME into MM surfaces by mean of magnetic attraction in a stirred solution; (2) Investigation if ME could reduce fouling on MM in comparison with a control; and, (3) Investigation of condition if ME could be used for enzymatic cleaning of fouled membranes. The experiment set-up used for this experiment is shown in Figure 2. The ability to magnetically attraction ME into MM was tested in a continuously stirred reactor. A MM module was placed inside the reactor, which contain solution in addition of ME. The attraction of ME to MM was observed visually. To investigate the effectiveness of the system to control fouling, the ME (190 AU) was added to a bulk solution containing 1 g/L arabinoxylan. In this case, arabinoxylan acted as substrate for ME and as solute to be filtered by the membrane. Meanwhile, the control reactor contained of NP and arabinoxylan. The filtration was derived from peristaltic pump and the TMP was measured from installed manometer. To obtain a moderate TMP built-up, the filtration flux was set at 6.58 L/m 2 /h, and the TMP was monitored over time (24 hours) for both the control and MEMM reactors. A mechanical stirrer was used to homogenously mix the solution. The magnetic field was derived from permanent magnet placed at top of both reactors. In the final phase, the cleaning potential of ME was also investigated. The fouled module in the control solution was brought into the reactor, which contained ME without cleaned. Meanwhile, the module in MEMM system was first physically cleaned to be sure all ME were removed from the membrane.

Membrane characterization
The permeability of the membranes was determined with the 'High throughput (HT)' apparatus (Vandezande et al., 2005) at constant pressure of 0.25 bar. With this device, the permeability of 16 samples (2.24 cm 2 ) cut from membrane sheets (n=5), could be obtained simultaneously. The flux (J) and permeability (L) of the membrane was calculated by using equations (1) and (2), respectively. Equations (1) and (2) can be expressed in the following: where J is the flux (L/m 2 /h), V is the volume (L), t is the time (hours), L is the permeability (L/m 2 /h/bar), and TMP is the transmembrane pressure (bar). The CWP was determined for fresh membranes without MNP and MM to compare their properties.
The microstructure of the membranes was observed using a scanning electron microscope (SEM; Philips SEM XL30 FEG with Adax dx-4i system). To obtain the thickness, the SEM images of the cross section of the membrane were visually observed and measured after breaking the membranes under a liquid nitrogen. To further analyze the presence of MNP in the MM, an elemental analysis to identify the composition of the inorganic specimen in the MM was performed by using an energydispersive X-ray spectroscopy (EDX) that is integrated with SEM.

Determination of the activity and activity unit
The activity of the BsXynA was determined by using the Colorimetric Xylazyme AX method according to Pollet et al. (Pollet, 2010;Pollet et al., 2009) The absorbance was plotted as a function of enzyme concentration and extrapolated to obtain the enzyme concentration where the absorbance should equal to 1.

High-Performance Anion-Exchange Chromatography (HPAEC) with integrated Pulsed Amperometric Detection (PAD)
The presence of coupling was analyzed using high-performance anion-exchange chromatography (HPAEC) with integrated pulsed amperometric detection (PAD) according to Pollet et al. (Pollet et al., 2009) In this test, the xylohexaose (X6) (Megazyme, Ireland) was used as a substrate because it has smaller structure. The L-Rhamnose (Sigma-Aldrich, Belgium) was used as a standard solution. The smaller substrate was used to prevent steric effect due to support material that probably existing in Xylazyme AX tablets. The experiments were performed in triplicate.

Fourier Transform InfraRed (FTIR) spectroscopy
The FTIR was used to identify the presence of -CONH-bond as a covalent bridge between enzyme and MNP. The immobilization of BsXynA occurred via amine groups present on the enzyme that bind to carboxylic groups on the MNP. As a consequence, the presence of a -CONHbond could prove the coupling of the enzyme to a MNP. Individual FTIR samples were prepared for NP and ME. Approximately 10 mg of dried NP and ME were mixed with potassium bromide (Fluka) in a proportion of 10/90 (%wt). The mixture was crushed into become powder and further pressed to form a thin transparent film.

Membrane and magnetic membrane characteristics
The effect of polymer concentration on CWP of fresh membrane and MM are shown in Figure 3. The permeability of the membranes, both fresh and MM were in the range of UF to MF. It decreased with the increase of polymer concentration. For MM, the CWP slightly decreased with increasing NP concentrations, and this is only significant for 30 wt% of MNP. The SEM images of the PSF and MM (surface and cross section) are shown in Figure 4. Results show that incorporation of MNP changed the membrane structure. The MM surfaces seem to be more compact in presence of MNP (20 and 30% for Figure  4(a) and (b), respectively). The white spots are clusters of MNP embedded in the PSF polymer matrix. The cross section images of magnetic membranes also show that the MM (38 µm) is thinner than the fresh membranes (90-100 µm), even though both were cast at similar wet thickness. The presence of MNP in the MM can be observed visually from the change of color from white in fresh membranes to brownblack in MM. The magnetic properties of MM were simply tested by putting them close to a permanent magnet. The MM attracted to the permanent magnet at the distance of 1-2 cm.  The presence of MNP in the MM was further analyzed by performing an elemental analysis using EDX. The elemental analysis of the MM is shown in Figure 5. Three elements were clearly present and dominant in the MM, namely sulfur (S) from polymer material (PSF), calcium (Ca) from tap water used to store the membrane, and iron (Fe) from MNP.

BsXynA activity
The activity of BsXynA was able to be measured using Colorimetric Xylazyme AX method (See Figure 6). Acording to the test, one XU was corresponded to 0.088 µg/mL of BsXynA. This value would be used to compare activities in the following experiments under different conditions. To obtain the optimum pH of BsXynA, relative activity of BsXynA was checked at different incubating pH as plotted in Figure 6(a). The highest activity was found at a pH of 6-7. Only less than 40% activity could be maintained when the incubation at pH between 5 and 8. The stability of the enzyme after an expose to different buffer solution for two hours, is shown in Figure  6(b). The relative activity of enzyme changed when it was stored not at their optimum pH. The loss of activity was tolerable for a pH range of 4 to 8 with less than 20% activity loss. However, too acidic or alkaline solutions could significantly reduce its activity. For instance, at pH 3, the enzyme lost 50% of its activity.
The relative activity of BsXynA at different temperatures is shown in Figure  6(c). The highest activity was found at 40 °C. Total loss of activity was found at 60°C and up to 80% loss when the reaction temperature down to 20°C. To explore the allowable range of enzyme storage temperature and coupling condition, the stability of enzyme after stored at different temperatures for 40 minutes was tested. The effect of storage temperature on enzyme activity is shown in Figure 6(d).
BsXynA was relatively stable at a temperature below 40°C. However, above it, significant activity loss was found. An increasing of temperature to above 50°C leads enzyme denaturation, resulting in a total loss of its activity. Further experiments for activity measurement were carried out in MOPS buffers (20 mM) (or McIlvaine buffer) at a pH of 6.5 and incubation temperature of 40°C.

Coupling analysis
The effectiveness of coupling of BsXynA into MNP using EDC/NHS as cross linking agent was observed using two methods: FTIR and HPAEC-PAD. The absorbance spectra of ME and NP from FTIR analysis are shown in Figure 7. The peaks below 1500 cm-1 represent organic content of the enzyme. The presence of carboxylic groups in NP and MEC was proven by spectras at wavelengths of 3000 -2500 cm-1. As prove of coupling, the presence of the -CONHbond is most important to see from the spectra. This bond is located in wavelengths of 1680 -1630 cm-1 and highlighted in Figure 7. The ME absorbance showed two peaks around 1669 and 1649 cm-1 which correspond to the presence of the -CONHbond which prove the coupling of enzyme to MNP.
The ME made from MNP with 10% acid groups and enzyme/MNP ratio of 0.1 to 1 was used as sample for HPAEC-PAD analysis. The chromatograms of the hydrolysis products produced from wild type xylanse and the ME are shown in Figure 8(a) and  (b). It is clear that xylohexaose was mainly cut into xylotriose both for wild type xylanse and the ME. Result indicates that ME was contained with BsXynA which eventually broke down xylohexaose into xylotriose as shown in Figure 8(b).

Coupling optimization
The surface area of peaks on the chromatograms from HPAEC-PAD analysis is correspond to the enzyme activity. Comparison of wild type xylanse to ME shows of only 0.583% of free enzyme activity can be maintained after the coupling process. The analysis using Colorimetric Xylazyme AX method was also found in similar magnitude. This could be due to loss of activity during the coupling process or due to low efficiency of coupling. To obtain high activity yield on the ME, a series of coupling with different operational parameters were performed to optimize the coupling: (1) percentage of acid groups on the MNP, (2) the ratio of enzyme to MNP, (3) cross-linking reaction time (step 2), (4) different buffers in the activation step (step 1) and reaction step (step 2) and (5) storage of MEC in different buffers. For coupling optimization, the activity was measured using Colorimetric Xylazyme AX method.

Effect of the ratio of acid group on hydroxyl group on the MNP
A series of coupling with different acid groups on MNP (0.1%, 1%, 5%, 10% and 100%) were performed using enzyme to NP ratio: 0.1 mg of enzyme/mg NP, 2 hours of reaction time. The activity of ME and the activity of uncoupled enzymes left in the supernatant were measured and the results are shown in Figures 9(a) and (b). The activity of ME and supernatant are represented as their relative activity compared to the initial amount of activity of BsXynA added to perform the coupling. Results show that ME formed from MNP with 10% of acid group has the highest activity. This MNP was used for further tests.  Figure 10 shows the activity of ME and non-coupled enzyme residue in supernatant for MNP with different percentages of acid groups. In the figure, the different scale of yaxis must be considered. In the experimental procedure, the coupling of enzyme to MNP was performed with three different enzyme/MNP ratios of 0.1, 1, and 10 and 2 hour coupling reaction. The activity of ME and the activity of uncoupled enzymes left in the supernatant are shown in Figures 10(a) and (b). The highest activity was found for enzyme/MNP ratios of 0.1. This ratio was used for further tests.

Effect of reaction time
To investigate the effect of the coupling reaction time (step 2), the coupling of enzyme to MNP were performed for 2 and 24 hours. The activity of coupled enzymes (ME) and the activity of uncoupled enzymes left in the supernatant are shown in Figures  11(a) and (b). The reaction times did not have a significant influence on the coupling efficiency of ME. The lower activity for longer reaction duration was probably due to longer storage time at room temperature during the reaction. The reaction time of 2 hours was used for further tests.  Figure 12 shows the activity of ME and non-coupled enzyme residue in supernatant with different buffer conditions in coupling steps. The effect of different buffer solutions in step 1 and 2 of the EDC/NHS coupling procedure was examined. All optimized parameters from previous test were used in this test, except that activity was measured in McIlvaine buffer instead of MOPS buffer.

Effect of different buffers in the activation step (step 1) and reaction step (step 2)
The activity of ME and supernatant are shown in Figures 12(a) and (b). The highest immobilization obtained based on activity measurements was for a combination of H 2 O in steps 1 and 2 in addition of few drops acetic acid at step 1. The coupling efficiency of 1.2% was obtained.

The effect of storage of MEC in MOPS buffer versus McIlvaine buffer
The effect of ME storage buffer after coupling was observed in this test. The ME was prepared according to the all optimum parameters obtained from previous tests. The activity of ME was compared for storage in McIlvaine or MOPS buffer. The relative activity of ME and supernatant in different storage solutions are shown in Figure 13(a) and (b). Results show that the activity of ME was higher when stored in MOPS (8.3%) compared to McIlvaine buffer (1.2%). The difference in residual activity is probably due to different separation efficiency. The ME seem to be more dissolve in McIlvaine than in MOPS buffer. McIlvaine buffer has good interaction with ME and keep ME remain in suspension. This is also confirmed by the residual activity measured in the supernatant of MOPS (26.1%) and McIlvaine (46.0%) buffer. The ME that present in McIlvaine, were probably washout with the supernatant in washing step due to a poor magnetic settling separation.

Magnetic attraction and MNP recovery
The preliminary system test was conducted to evaluate the ability of a super paramagnetic material (in this case, weak iron) to magnetically attract ME from a wellmixed-bulk solution. In this case, weak-iron mimicked the function of MM. The result of the system test is shown in Figure 14. When the system induced with magnetic field by placing the permanent magnet close to the weak iron, the suspended ME was attracted into the surface of weak-iron. However, the ME were not distributed homogeneously over the surface. This is because of magnetic strength distribution. This phenomenom is a typical characteristic of magnetic materials. They posses stronger magnetic properties on their edges than in the middle. However, the test using MM module revealed that the ME was attracted well distributedly, since MNP were distributed into the polymer membrane matrixes (pictures not shown). This first phase of the system test proved that ME could effectively be magnetically attracted onto a super paramagnetic surface.

Fouling reduction potential of ME
The fouling reduction potential by means of in-situ enzymatic cleaning of membranes was investigated in a dead-end filtration system. The MM was potted into two small envelopes which have an effective filtration area of 2 x (4 x 8) cm2. One membrane was used in presence of ME and the other as a control in presence of MNP. The TMP profile of the filtration is shown in Figure 15.

ME-MM
For both reactors, the TMP rose exponentially at the beginning (up to 3 hours) of filtration, due to polarization concentration and rapid foulant build-up at the membrane surfaces. After that, a continuous exponential increase of TMP occurred for control reactor. The back transport from hydrodynamic created by mechanical mixing was lower than the fouling rate resulting the severe fouling. The TMP start to rise at lower rate for the following stage, reaching 243 mbar after 24 hours of operation. In the MEMM system, after initial exponential rose up to 3 hours from the starting the filtration, the TMP started to stabilize at a much lower TMP of 47 mbar, following by weak rise of TMP and reached a final value of 97 mbar after 24 hours of operation. After subsequent amount of foulant accumulated at membrane surface, the foulant degraded by ME facilitating the prevention of exponential rise of TMP. These results showed a preliminary indication on the effectiveness of the MEMM system to control membrane fouling. The ME in bulk attracted by mean of magnetic forces onto MM surfaces and work very locally, precisely where it required on the membrane surface. This makes MEMM system is expected to be a very efficient for fouling control. However, by the time passed the lost on the ME activity occurred resulting the slow increase of TMP.

Enzymatic cleaning potential of MEC
After the test of Section 4.4.2 had finished, the membranes from both systems (MEMM and control) were switched. The MM (M-1) that was placed in the MEMM system was switched with the MM (M-2) that was placed in the control system. Prior to the filtration started, M-1 was physically cleaned by using pressurized tap water to release the irreversibly attached ME from the surfaces and M-2 left in fouled condition. The profile of filtration for the test on the cleaning potential of MEMM system is shown in Figure 16. As expected, the initial TMP of M-2 is higher than M-1. In case of M-1, considerably high initial TMP was due to the foulant trapped inside the pores that could not be removed and for M-2 it was obviously due to the accumulated foulant from previous test. After an extension of filtration period, the TMP increased at an almost exponential rate during the first two hours and attained a value of 207 mbar after 24 hours for M-1. A different TMP behavior was obtained for M2. The TMP first decreased for the first three hours and reached a value of 153 mbar following by weak increase of TMP. The decrease of TMP just after the filtration started indicates the occurrence of enzymatic cleaning on the fouled membrane surfaces. However, the TMP of M-1 started to increase again due to a loss in ME activity but still showed better performance than the control (M-2).