An enzymatic membrane reactor for oligodextran production: Effects of enzyme immobilization strategies on dextranase activity

An enzymatic membrane reactor (EMR) with immobilized dextranase provides an excellent opportunity for tailoring the molecular weight (Mw) of oligodextran to significantly improve product quality. However, a highly efficient EMR for oligodextran production is still lacking and the effect of enzyme immobilization strategy on dextranase hydrolysis behavior has not been studied yet. In this work, a functional layer of polydopamine (PDA) or nanoparticles made of tannic acid (TA) and hydrolysable 3-amino-propyltriethoxysilane (APTES) was first coated on commercial membranes. Then cross-linked dextranase or non-cross-linked dextranase was loaded onto the modified membranes using incubation mode or fouling-induced mode. The fouling-induced mode was a promising enzyme immobilization strategy on the membrane surface due to its higher enzyme loading and activity. Moreover, unlike the non-cross-linked dextranase that exhibited a normal endo-hydrolysis pattern, we surprisingly found that the cross-linked dextranase loaded on the PDA modified surface exerted an exo-hydrolysis pattern, possibly due to mass transfer limitations. Such alteration of hydrolysis pattern has rarely been reported before. Based on the hydrolysis behavior of the immobilized dextranase in different EMRs, we propose potential applications for the oligodextran products. This study presents a unique perspective on the relation between the enzyme immobilization process and the immobilized enzyme hydrolysis behavior, and thus opens up a variety of possibilities for the design of a high-performance EMR.


Introduction
The enzymatic membrane reactor (EMR) is nowadays regarded as a green platform that enables the integration of bioconversion and membrane separation (Giorno et al., 2014;Giorno & Drioli, 2000). The EMR approach, in which the enzymes function as efficient biocatalysts in concert with a membrane separator for simultaneous product purification, has been increasingly reported for its various applications in both upstream and downstream processes (Jochems et al., 2011;Luo et al., 2020). One of the most significant applications of the EMR is the production of oligosaccharideslow molecular weight (Mw) carbohydrates with the number of sugar monomers intermediate of simple sugars and polysaccharideswhich have high commercial value due to their specific chemical structures and unique physicochemical properties (Zhao et al., 2021). With increasing demand for oligosaccharides on the global market, the production of oligosaccharides not only requires environmentally friendly processes but also a smart technology for precise control of product Mw during fabrication. The EMR is no doubt one of the ideal options for meeting both demands.
Traditional production of oligosaccharides introduces a considerable amount of hazardous chemicals, which potentially cause immune risks in practical usage of the products Su et al., 2020). To address the undesired issues in production, our previous study used dextranase to convert polydextran to oligodextran while a membrane simultaneously functioned as a selective sieve to obtain the intermediate Mw oligodextran products (Su et al., 2018). The abovementioned work provided a strategy to tailor the Mw of oligedextran and thereby increase the product quality. Moreover, to obtain maximum amount of the target oligodextran products, the enzymatic hydrolysis should occur near the membrane surface for immediate removal of the target oligodextran from the reaction system and to avoid over-degradation. By this approach, products with narrow Mw distribution could be obtained. Enzyme immobilization on the membrane therefore offers a promising opportunity for better control of the overall process near the membrane surface.
Membrane modification is commonly carried out to make the membrane susceptible to enzyme immobilization (Qing et al., 2019). Polydopamine (PDA), a neurotransmitter that easily forms a thin coating layer by self-polymerization in alkaline aqueous solution, is reported to serve as a functional layer that enables the conjunction of enzymes and exposed catechol and quinone groups of the PDA layer (Alfieri et al., 2018). Based on the above theory, Zhang et al. established a versatile PDA coated membrane platform onto which dextranase was covalently attached (Zhang et al., 2018). Besides providing functional groups for the stable attachment of enzymes, the PDA coating improves the hydrophilicity of the membrane substrate, which contributes to increase the water permeability (Fan et al., 2017). In An alternative approach, Wang et al. (2018) developed a hierarchical coating layer on a membrane surface based on the secondary reaction between tannic acid (TA) and hydrolysable 3-amino-propyltriethoxysilane (APTES). The hierarchical TA/APTES nanosphere layer, which is rich in quinone groups, provides a hydrophilic, functional surface to which enzymes can readily attach . Zhou et al. (2020) further investigated the effect of the TA/APTES ratio on the enzyme loading efficiency and found that the enzyme loading could be greatly increased via TA/APTES surface modification, notably due to the occurrence of abundant quinone groups on the surface as well as the vast increase in surface area following the formation of the TA/APTES nanospheres.
Following membrane modification, glutaraldehyde (GA) is often introduced to form covalent bonds between the enzymes and the coating layer (Sigurdardóttir et al., 2018). The high activity between aldehyde groups on the coating layer and amine groups on the enzymes enables a high enzyme loading efficiency (Barbosa et al., 2014). Moreover, the GA molecules can easily react with the amino groups on different enzymes to form cross-linked enzymes aggregates (CLEAs). CLEAs are reported to maintain high enzyme stability and have therefore attracted considerable attention in commercial applications (Sheldon, 2007). Enzyme loading efficiency is also affected by the mode of immobilization. Incubating the modified membrane in enzyme solution is the most common immobilization strategy but in incubation mode, enzyme loading efficiency is often hampered by mass transfer limitations (Rana & Matsuura, 2010). Thus, the driving force of enzymes moving towards the modified membrane surface needs to be enhanced to improve the enzyme loading efficiency. A fouling-induced method, inspired by the mechanism of membrane fouling, has been proposed as a promising strategy to enhance enzyme concentration near the membrane surface (Luo et al., 2013;Morthensen et al., 2017). The enzyme immobilization strategies described above provide various possibilities for the design of an EMR. In this study, we evaluated two membrane surface modification methods and two enzyme immobilization methods for the immobilization of dextranase on ultrafiltration (UF) membrane substrates. Thus, we coated the membrane substrates with either PDA or TA/APTES, followed by immobilization of dextranase via incubation or fouling-induced mode. Subsequently, we evaluated the respective strategies based on their performance in terms of production of oligodextran. Previous studies on dextranase immobilization have aimed at optimizing the hydrolysis rate of the enzymes (Bertrand et al., 2014;Shahid et al., 2019) but lack a discussion of tailoring the enzyme hydrolysis behavior to control the Mw of oligdextran. Therefore, besides focusing only on high enzyme loading and high enzyme activity retention upon immobilization, we also investigated the effects of the different immobilization strategies on the catalytic behavior of immobilized dextranase and compared the corresponding enzyme activity. Gel permeation chromatography (GPC) was used to analyze the components of the hydrolyzed oligodextran products in different EMRs, which illustrate the different hydrolysis patterns of the immobilized dextranase. Based on the hydrolysis patterns of the immobilized dextranase, future applications of different enzyme immobilization strategies are proposed. Our work indicates multiple possibilities for the design of a high-performance EMR.
Membrane modification with either dopamine or TA/APTES, enzyme immobilization and activity assay of immobilized enzymes were performed in a stirred cell (Amicon 8050, Millipore, USA) with an effective membrane surface area of 13.4 cm 2 .

Enzymatic membrane preparation by different immobilization strategies 2.2.1. Membrane modification
Dopamine or TA/APTES mixture was applied for surface modification of pristine commercial membranes. For dopamine modification, pristine membranes were incubated with 10 mL of 2 g/L or 4 g/L dopamine hydrochloride solution (pH 8.5, 10 mM Tris-HCl buffer) at 100 rpm and 25 • C for different time-periods (1 h, 2 h or 4 h). Membrane modification by TA/APTES was carried out according to the work of Zhou et al. (2020): briefly, 2 g/L TA solution in Tris-HCl buffer (pH 8.5) was mixed with a 10 g/L APTES in EtOH solution at a volume ratio of TA/APTES = 8:1 to make 20 mL coating solution. Pristine membranes were then incubated in the TA/APTES coating solution at 100 rpm and room temperature (25 • C) for 18 h. The TA/APTES modification introduced a layer of nanospheres on the membrane surface that is rich in quinone groups for enzyme immobilization by covalent bonding. After modification, the membranes were cleaned using running distilled water to remove the residual modifiers and then the modified membranes were installed into the Amicon cells for enzyme immobilization.

Enzyme immobilization
Enzyme immobilization on dopamine or TA/APTES modified membranes was carried out in incubation mode and fouling-induced mode. With dopamine modified membranes, 10 mL of 2 g/L dextranase solution (with 605-668 μg soluble proteins) containing 1% (v/v) GA was placed in contact with the membrane surface in the Amicon cell. In the incubation mode, the enzyme solution was incubated with the membrane for 2.5 h at 100 rpm, after which the enzyme solution was recovered from the Amicon cell and stored for protein concentration measurements by Bradford assay. In the fouling-induced mode, the enzyme solution was incubated with the membrane for 1 h at 100 rpm, and then the enzyme solution was filtered at 0.2 bar until all the solution was permeated from the cell. The permeate was collected for protein concentration measurements.
With the TA/APTES modified membranes, the enzyme immobilization occurred through covalent bonding between amino groups on the enzymes and the quinone groups on the coating layer, which formed via Michael addition and Schiff's base reaction. In the incubation mode, the enzyme solution (10 mL of 2 g/L dextranase) was added to the Amicon cell and the membrane was incubated for 2.5 h at 100 rpm. The enzyme solution was recovered from the cell after the immobilization for protein concentration measurements. In the fouling-induced mode, 10 mL of 2 g/L dextranase solution was filtered at 4 bar and 500 rpm until all the solution was permeated from the cell. The permeate was collected for protein concentration measurements. After enzyme immobilization, each membrane was washed three times with 5 mL of pure water.
Enzyme immobilization experiments performed by the four independent methods were conducted in duplicates.

Enzyme loading determination
The protein concentration of the enzyme solutions was measured by the modified Bradford assay according to  0-16 μg/mL of bovine serum albumin (BSA) solutions were used for the calibration. Samples were diluted to be within the range of the protein calibration curve, as required. The enzyme solutions were mixed with Bradford reagent in a 1:1 volumetric ratio. After 5 min of incubation, absorbance was measured at 595 nm. Enzyme loading mass was calculated from the equation: where c is the soluble protein concentration and V is the volume of the solution at the corresponding concentration. Subscripts i, r, p and w represent initial, recovered, permeate and washing solutions, respectively. The enzyme loading is defined as:

Enzyme activity determination 2.3.1. Activity of immobilized and free enzymes
To measure the observed activity of the immobilized enzymes, 20 mL 4 g/L DXT70K solution was added to a 50 mL Amicon stirred cell (Amicon UFSC05001, Merck Millipore, USA) with the enzymatic membrane at room temperature and 100 rpm. Samples were collected at specified time intervals. To measure the activity of free enzymes, 1 mL of 2 g/L dextranase solution (or dextranase solution with 1% v/v GA) was introduced into 20 mL 4 g/L DXT70K solution for 90 min. Samples were collected every 5 min, then incubated in a boiling water bath to fully stop the reaction at specified time points. The reducing sugar content of all the collected samples was measured by using 3, 5-dinitrosalicylic (DNS) acid reagent, according to the method modified by Zhang et al. (2018). Specifically, 1 mL hydrolyzed samples were mixed with 1 mL DNS reagent and heated in a boiling water bath for 5 min. The samples were diluted 5 times by ultrapure water and measured at 540 nm. Immobilization yield, efficiency and activity recovery were calculated from the following equations (Sheldon & van Pelt, 2013): Activity recovery(%) = observed activity starting activity × 100% The immobilized activity was determined by measuring the total residual enzyme activity after immobilization and by subtracting this activity from the total starting activity. The enzyme activity was defined as the amount of isomaltose (measured in μmol maltose) generated after 1 min at 25 • C, using μmol-isomaltose/min units. The enzyme activity tests of starting solution, residual solution and the immobilized dextranase were tested at 25 • C in duplicates.
The average Mw of the above samples was later tested in a Thermo Scientific -GPC system.

Enzyme kinetic parameter measurement
The Michaelis− Menten kinetic parameters K m and V max of enzymes were determined by measuring the initial rates of the catalytic reactions using different substrates.

Membrane surface morphology
Scanning electron microscopy (SEM) was used to visualize the morphology of PDA modified PES membranes with immobilized enzymes. Here, samples with gold coating (Balzers PV205P, Switzerland) were investigated using an EVO40 microscope (Zeiss, Germany).

Effect of enzyme immobilization mode on enzyme loading
Firstly, the effects of PDA coating parameters on enzyme loading were investigated (Table S1), and it was found that neither increased PDA concentration nor coating time significantly improved enzyme loading in incubation mode. A possible explanation is that the PDA layer might tend to form a brush-like surface that prevents the attachment of enzymes (Gao et al., 2011;Cai et al., 2012).
To improve enzyme loading efficiency on the membranes, we investigated methods to overcome the repulsion between the enzymes and the membrane coating layer. More enzyme-membrane contact could be achieved either by increasing the initial enzyme concentration or by applying pressure above the membrane. The latter strategy is known as fouling-induced enzyme immobilization. This method uses pressure to increase the enzyme concentration near the membrane surface (i.e. concentration polarization) (Luo et al., 2014). In the following study two different enzyme loading modesincubation mode and fouling-induced modewere compared.
The fouling-induced mode was applied to increase the enzyme loading efficiency on the PDA coated membrane surface. Fig. 1A illustrates the enzyme distribution on membranes prepared using two different immobilization modes. 49% (326.7 μg) of dextranase (in terms of protein mass) was found on the membrane surface when the foulinginduced immobilization mode was applied, whereas only 16% (107.8 μg) dextranase was loaded on the membrane surface in incubation mode. The proposed mechanisms are shown in Fig. 1B where GA forms covalent bonds between the enzymes and the PDA layer and simultaneously functions as an enzyme cross-linker to form CLEAs. In Fig. 1C and D. The CLEAs measured over 1000 nm in size, while the PDA particles (bright circles) had a diameter around 50 nm, which is similar to results reported by (Li et al., 2014). The coating layer weakened the total interaction (a sum of acid-base (AB), Lifshitz-van der Waals forces (LW) and electrostatic double layer interactions) between the enzyme aggregates and the modified membrane (Cai et al., 2017), which could result in most of the dextranase (81%) remaining in the solution after 2.5 h incubation. In the fouling-induced mode, however, the enzymes together with GA were filtrated towards the membrane surface by convective transport when the solvent passed through the membrane. From the perspective of adhesion energy, the strong driving force due to the filtration might overcome the static repulsion between the rough coating layer and the CLEAs. Under these circumstances the enzymes would not diffuse back to the bulk solution, but would instead contribute to an increase in local concentration at the membrane surface, with more efficient covalent bonding between enzyme and the membrane as the result. Consequently, a higher enzyme loading would be obtained on the membrane surface in fouling-induced mode than in incubation mode.
The result indicates that the applied pressure provides a driving force that overcomes the steric hindrance between enzyme clusters and the PDA coating, resulting in a higher enzyme loading efficiency.
The enzyme activity of the catalytic membranes was evaluated for 1260 min (21h) to observe the degradation efficiency of the immobilized dextranase (Fig. 2). With a higher enzyme loading, the enzymatic membrane in fouling-induced mode showed increasing activity within the first 120 min. Over the same reaction period (Fig. 2B), a rapid decline of dextran Mw was observed. By contrast, the enzyme activity in the incubation mode was low, and consequently, the accumulation of reducing sugars within the first 60 min was slow. Therefore, the observed peaks of isomaltose were not as obvious compared with the bulk dextran substrate (Fig. S1). In incubation mode, in accordance with the low activity, the decrease of dextran Mw was slow. Regarding the enzyme hydrolysis efficiency, the dextranase immobilized in foulinginduced mode outperformed those immobilized in incubation mode and led to a faster degradation of large dextran molecules.
Interestingly, when investigating the composition of the hydrolyzed oligodextran products in detail ( Fig. 2C and D), the immobilized enzymes introduced by the different modes were found to have different hydrolyzing patterns. The dextranase (from Penicillium sp.) used in this study is reported to be an endo-glycosidic enzyme that randomly attacks the α-1,6 glycosidic bonds within large dextran molecules and releases shorter oligodextran until the hydrolyzed products become dimers. By contrast, exo-glycosidic enzymes degrade the dextran chains from the terminal side of the molecule to release end-products such as dimers or monomers (Khalikova et al., 2005). The GPC chromatograms in our study show that dextranase immobilized by incubation mode tended to produce end-products (single units of isomaltose) during the reaction and that the bulk of the large dextran molecules remained unattacked at the beginning. This finding indicates that the dextranase immobilized in incubation mode performed exo-hydrolysis so that products with a very broad Mw distribution were produced. By contrast, there was an overall Mw decline of the bulk dextran molecules on the membrane with fouling-induced enzymes while accumulation of end-products occurred during the hydrolysis. The results suggest that part of the foulinginduced dextranase on the membrane surface maintained the endohydrolysis pattern. Such a shift in hydrolysis performance of the immobilized dextranase has rarely been reported.
Immobilization efficiency, activity recovery, and immobilization yield are indicated in Table 1. The fouling-induced mode yielded a significantly higher immobilization yield (75.2%), efficiency (6.7%) and activity recovery (5.0%) compared to the corresponding values of the incubation mode (44.5%, 2.1% and 0.9%, respectively). Shahid et al. (2019) reported similar immobilization yield (34%-78%) of dextranase immobilized on an alginate matrix. The low activity recovery is due to a relatively large enzyme amount at the starting solution (605-668 μg soluble proteins) and to the limited membrane surface that did not allow more enzymes to be immobilized. Secondly, the dextran macromolecules cannot easily penetrate the CLEAs, which leads to an activity Table 1 Enzyme immobilization efficiency, activity recovery and immobilization yield of the cross-linked dextranase on the PDA modified membrane. a The total enzyme activity on membrane is calculated by total enzyme activity in starting solution subtract total enzyme activity in residual solution after immobilization; the observed enzyme activity on membrane was measured by terminology mentioned in Section 2.3.1.

Carbohydrate Polymers 271 (2021) 118430
6 decline of the immobilized enzymes (Sheldon et al., 2021). Though only 5% of the initial enzyme activity was recovered in our work, the immobilized enzymes gradually catalyzed the dextran substrates into oligodextran products (Fig. 2B). The slower reaction enabled Mw tailoring during production, which offers a promising application for the EMR.
The fouling-induced mode exhibited higher enzyme immobilization efficiency and activity, which potentially could be applied at a larger scale to increase oligosaccharides productivity. However, during the 21 h enzyme activity test, around 10% of the immobilized enzymes in fouling-induced mode (30 μg) leaked from the membrane surface, whereas no enzyme leakage was detected in the incubation mode. With incubation mode, in this regard, most enzymes were firmly immobilized via covalent bonding, which is beneficial for long-term usage due to reduced loss of enzyme to the surrounding environment.

Enzyme immobilization on TA/APTES modified membrane surfaces
Besides PDA, we applied TA/APTES surface modification for the immobilization of dextranase. Here, GA was not mixed with the dextranase because the reported results (in supplemental information Section B.2.4) suggested that non-cross-linked dextranase might perform better in hydrolysis. Applying similar procedures as described earlier, dextranase was immobilized using incubation and foulinginduced modes, respectively, on the TA/APTES coated membranes. The observed higher enzyme activity (Fig. 3) suggests a higher enzyme loading on the TA/APTES modified membrane surface in comparison to the PDA modified membrane surface. Moreover, according to a report from , the TA/APTES nanoparticles have an average diameter of around 200 nm. Obviously, the spherical nanoparticles formed by TA/APTES are larger than the PDA particles (<50 nm) and form a larger asperity radius on the coating layer. The reported molecular size of dextranase from Penicillium sp. is around 67 kD (Larsson et al., 2003). Based on this molecular size, the diameter of a single and isolated enzyme molecule is estimated to be around 3.3 nm according to the protein size approximation from Erickson (2009) which assumes the enzyme has a spherical shape. Due to the small size of noncross-linked enzymes and relatively larger size of the TA/APTES particles, the TA/APTES coating layer is expected to provide a larger surface area on the membrane available to dextranase to anchor on. In the perspective of thermodynamics, the interaction energy between proteins and the membrane surface increases with a larger asperity radius of the membrane (Zhao et al., 2015). That is, enzyme adhesion on a larger particle (TA/APTES) surface should have a higher entropy .
Regarding enzyme immobilization modes, as expected, the enzymes introduced by fouling-induced mode showed a relatively high enzyme activity (ca. 0.7 μmol-isomaltose/min) particularly within the first 60 min of hydrolysis reaction. Correspondingly, the immobilized enzymes efficiently converted long chain dextran substrates into smaller units resulting in a rapid Mw decrease of dextran at the beginning of reaction (Fig. 3B). In the incubation mode, much lower enzyme activity was observed, leading to a slower dextran Mw decline. In a similar manner to the explanation discussed in Section 3.1, strong convection in the fouling-induced mode brought more enzymes towards the membrane surface, which enabled a higher enzyme immobilization efficiency and higher activity.
Though the enzymes in the incubation mode showed a slower overall Mw decline, not much accumulated end-products (separated peak of isomaltose) were observed from the GPC chromatograms. The results indicate that the non-cross-linked enzymes attached to the TA/ATPES surface exerted the desired endo-hydrolysis when producing the oligodextran. When looking into the product composition generated by the fouling-induced enzymes, a sharp decline of the overall Mw of dextran substrates was observed at the beginning of the reaction (Fig. 3B), however without an immediate accumulation of end-products (i.e. isomaltose, Fig. 3D). The rate of end-product accumulation indicated that the appearance of end-products was due to an efficient degradation of dextran instead of due to exo-hydrolysis. With larger capacity of immobilized enzymes, the dextran substrate (70 kDa) was efficiently hydrolyzed into small units (around 8.9 kDa) even within the first 15 min reaction.
Additionally, enzyme activity loss was observed on the EMR after reusing the biocatalytic membranes in several reaction cycles (Fig. S3). Though covalent bonding is targeted, most of the enzymes loaded using fouling-induced mode are immobilized by adsorption due to the strong convection. The extent of leakage of immobilized enzymes from the support during the repeated cycles is similar to the one reported by da Silva et al. (2019).

Comparison of enzyme immobilization strategies
The previous sections indicate that EMRs based on different strategies exert different degradation behaviors (Fig. 4). Regardless of the modifiers, EMRs based on fouling-induced mode exhibited a significantly higher dextran Mw decreasing rate at the beginning of hydrolysis. On the other hand, TA/APTES coated membranes seemed to retain higher enzyme activity. Based on the analysis above, the TA/APTES layer provides a larger surface for enzymes to attach to and hence more dextranase is supposedly loaded onto its surface.
Another important factor affecting the activity of the immobilized enzyme is the CLEAs triggered by GA molecules (Migneault et al., 2004). The tight packing of the cross-linked dextranase might cause severe mass transfer issues when hydrolyzing the dextran substrates, and thus lower degradation rates (Verma et al., 2019). According to many studies of enzyme immobilization, enzyme activity decline due to aggregation is commonly observed (Nadar et al., 2016). Therefore, without GA molecules, the non-cross-linked dextranase on the TA/APTES should contribute to the high dextran Mw decrease rate. Enzyme kinetic studies may explain why the enzyme activity varies in different EMRs.

Enzyme kinetics
To better understand the hydrolysis efficiency of cross-linked and non-cross-linked dextranase, the effect of addition of GA cross-linkers on dextranase hydrolysis behavior was studied under room temperature (25 • C). In this experiment, the enzymes were not loaded onto a membrane but were directly mixed with the substrates. Kinetic parameters are given in Table 2. Higher V max was observed in the reaction without GA cross-linked dextranase. The reason could be due to the location of the active center in the middle of the enzyme molecules (Zhang et al., 2018), which makes a part of the active sites inaccessible to the substrates upon cross-linking, thereby yielding a lower reaction velocity. Similar K m values indicate similar affinity between substrates and the non-cross-linked dextranase and GA-cross-linked dextranase. Similar V max and K m change of the immobilized dextranase are found in research from El-Tanash et al. (2011). However, the Michaelis-Menten parameters only describe the reaction velocity of the enzyme at the beginning of the hydrolysis reaction (Ivanauskas et al., 2016;Johnson & Goody, 2011). The hydrolysis behavior over an extended period should also be investigated because, in real applications, the enzymes are generally expected to perform the hydrolysis during an extended run time. Fig. 5 shows the specific enzyme activity of the GA-cross-linked and non-cross-linked dextranase during 90 min hydrolysis. The non-crosslinked dextranase gave very high specific enzyme activity in the beginning of the reaction, while the cross-linked dextranase showed lower and more steady specific enzyme activity. Thus the dextran substrates were almost fully degraded by the non-cross-linked dextranase within the first 5 min of reaction and the cross-linked dextranase resulted in a gradual Mw decline of the substrates (Fig. S6). When examined at the nanoscale, it has been reported that the structure of dextranase must change and form a tunnel-like space accommodating   both water molecules and dextran substrate so that the necessary nucleophilic attack can occur to cleave the α-1,6 glycosidic bonds within the large dextran molecules. (Larsson et al., 2003). The aggregated dextranase might, however, be limited by steric hindrance so that it takes longer for this aggregated dextranase to change its structures for the degradation of dextran substrates. Hence, intermediate Mw oligodextran was observed during the reaction. The delayed hydrolysis behavior may actually offer a possibility for better tailoring the production of oligodextran because these intermediate sized products are desired for their particular bioactive functions (da Silva et al., 2019;Rastall, 2010). In an EMR that integrates both bioreaction and separation processes, slower and controllable degradation of substrate Mw that matches the removal rate of the target molecules would help in improving the quality of the products (Su et al., 2018). On the other hand, the non-aggregated dextranase exerts high activity, so they are usually used to tackle substrates at higher concentrations and scales (Su et al., 2020).
Regarding mass transfer, the scenarios of large dextran molecules accessing the enzyme active sites on the membrane surface could vary with different EMRs that lead to exo-or endo-hydrolysis (Fig. 6). Fixed on the membrane surface, the non-cross-linked dextranase should have more exposed active sites facing the bulk solution, whereas many active sites of the aggregated dextranase are expected to be shielded. The dextran substrate in this study has an average Mw of around 70 kDa, and a hydrodynamic radii of around 9 nm according to the Stokes-Einstein relationship (Ioan et al., 2001). With the fixed enzymes on the membrane surface, the hydrolysis is assumed to be dominated by the Brownian movement of dextran substrates (Blanco et al., 2017). On the TA/ APTES coated surface, there should be larger spaces between the noncross linked dextranase that allows random Brownian movement of the dextran molecules. More effective collisions between the substrates and exposed active sites would therefore have occurred during the hydrolysis, resulting in an efficient attachment of substrates and quick detachment of products. Due to less steric hindrance between the enzymes and substrates, the immobilized dextranase could efficiently hydrolyze the dextran molecules (Frazier et al., 1997) exerting endohydrolysis that randomly attack the glycosidic bonds within the substrate molecules (Khalikova et al., 2005). However, the scenario around the CLEAs above the PDA coating layer might be different. The crosslinking led to high enzyme concentration in the CLEAs, and static compaction occurred among the enzymes, so there might be very limited space for whole dextran substrate to diffuse into the active sites for endohydrolysis (El-Tanash et al., 2011;. Instead, the limited space might only allow the terminal side of the large molecules to penetrate the CLEAs, leading to an exo-hydrolysis. Additionally, due to the larger size of CLEAs that have more binding sites for the substrates, once a large dextran molecule was attached on the CLEAs, it could hardly diffuse back to the bulk solution (Erhardt & Jördening, 2007), so the exposed active sites would continuously cleave the molecules until the smallest units, leading to the accumulation of endproducts during the intermediate process. Exo-hydrolysis thus has potential for the fabrication of low Mw oligosaccharides such as isomaltose (Zhou et al., 2019). The above analysis provides new insights into the working pattern of immobilized dextranase. With regard to specific products, the current study offers various selections of enzyme immobilization strategies.

Filtration performance of the EMR
We also evaluated the filtration performance of the membranes in terms of water permeability (Table S4). Both membrane modification and enzyme immobilization introduced extra filtration resistance to the membranes, which limits their separation performance in real processes. Therefore, the functional modification in combination with foulinginduced enzyme immobilization is proposed for application on a porous matrix, such as electrospun nanofibers . The enzymatic matrix could then be coupled with a membrane for product separation. Due to the loose structure of the matrix, the enzymatic layer would not introduce much filtration  High enzyme activity Hard to control the reaction Upscaling -handling substrates at a higher concentration that increase total productivity resistance above the membrane, and would therefore allow simultaneous enzyme reaction and products separation.

Conclusion
The current study evaluated different strategies of enzyme immobilization, each of which displayed different enzyme loading efficiency and activity. More interestingly, the type of catalytic activity of the immobilized enzymes was affected by the immobilization strategies.
The membranes coated by PDA or TA/APTES nanoparticles exhibited different surface morphologies and therefore different binding affinity to the enzymes. The 'fouling-induced' enzyme immobilization mode resulted in a higher enzyme activity, which therefore was adopted for a high-performance EMR. Furthermore, the enzyme kinetics of aggregated dextranase and non-aggregated dextranase was tested. Due to the GA cross-linker, the aggregated dextranase performed exo-hydrolysis on the membrane surface due to mass transfer limitations within the aggregated enzyme clusters. The filtration performance of the EMRs was compared to identify future applications of the EMRs. The above three factorsmodifier, enzyme immobilization mode and enzyme aggregation -are summarized in Table 3.
This work focused on the effects of enzyme immobilization strategies on dextranase hydrolysis behavior, and presents an in-depth discussion on the corresponding mechanisms. The results suggest various possibilities for the design of a high-performance EMR for the production of oligosaccharides.

Declaration of competing interest
The authors report no declarations of interest.