Production of Hypoallergenic Antibacterial Peptides from Defatted Soybean Meal in Membrane Bioreactor : A Bioprocess Engineering Study with Comprehensive Product Characterization

Arij it Nath, Gábor Szécsi, Barbara Csehi, Zsuzsa Mednyánszky, Gabriella Kiskó, Éva Bányai, Mihály Dernovics and András Koris* Department of Food Engineering, Faculty of Food Science, Szent István University, Ménesi st 44, HU-1118 Budapest, Hungary Department of Refrigeration and Livestock Product Technology, Faculty of Food Science, Szent István University, Ménesi st 43–45, HU-1118 Budapest, Hungary Department of Food Chemistry and Nutrition, Faculty of Food Science, Szent István University, Budapest, Somlói st 14–16, HU-1118 Budapest, Hungary Department of Food Microbiology and Biotechnology, Faculty of Food Science, Szent István University, Budapest, Somlói st 14–16, HU-1118 Budapest, Hungary Department of Applied Chemistry, Faculty of Food Science, Szent István University, Budapest, Villányi st 29–33, HU-1118 Budapest, Hungary


Introduction
Leguminous soybean (Glycine max) is considered the largest edible protein source around the world (1).Aft er wet milling of soybean seed, the main by-product soybean meal is generated, which is an abundant by-product of vegetable oil processing industries (2).Soybean meal contains 40-45 % (on mass basis) protein and its application is mostly limited to broiler and fi sh feed (3,4).It is well documented that functionality of plant proteins, including soy proteins, depends on amino acid sequence, their composition and digestibility.Although trypsin inhibitor activity is adequately reduced by the heat treatment during the production of soybean meal, the allergenic eff ect of soybean meal proteins renders this by -product an inadequate source of essential amino acids (5).To avoid this problem, production of hypoallergenic low--molecular-mass (LMM) peptides (protein hydroly sate) from defatt ed soybean meal is considered a reliable approach (6,7).
In many cases, acid hydrolysis is used to reduce the content of soybean allergens (8,9).For food-grade peptide production, concentrated hydrochloric acid treatment is not acceptable because of the formation of toxic compounds, such as 3-chloro-1,2-propane-diol and 1,3-dichloro-2-propanol (10).Therefore, enzymatic digestion is preferred because it can be controlled by the selection of specifi c enzymes and reaction conditions.Moreover, enzymatic digestion upgrades several functional properties and it can reduce the allergenicity of protein-based products (11,12).However, it must be noted that enzymatic treatment alone cannot defi nitely reduce the allergenicity of soy proteins (13).
Conventionally, batch bioreactors have been used for the hydrolysis of plant proteins, including soy protein (14)(15)(16)(17)(18)(19).Recently, some research groups have reported the application of membrane-integrated bioreactor for the production of angiotensin I-converting enzyme inhibitory peptides (20), antioxidant peptides (21,22) and anti-adipogenic peptides (23)(24)(25) from soy protein.To implement the membrane-integrated bioreactor on industrial scale for the production of bioactive compounds from soybean, it is clear that more systematic and judicious investigations with technological viewpoints are required.Methodical experiments with bench scale setup are a prerequisite to transfer the laboratory-developed technology to industrial scale.In our investigation, hypoallergenic antibacterial LMM peptides were produced from defatt ed soybean meal with an in-house developed membrane bioreactor, which is the fi rst such att empt to the best of our knowledge.Several advantages of the membrane bioreactor are well documented, e.g. it is effi cient for the separation of target product molecule and enzyme from the reaction broth and it also allows the reuse of the enzyme (26).
In the fi rst step of our experiments, defatt ed soybean meal was subjected to tryptic digestion in the bioreactor, operated in batch mode.Initial soybean meal concentration (γsm 0 ), reaction temperature and pH were varied under fi xed agitation speed to optimize the enzymatic reaction.Aft er the tryptic digestion, LMM peptides were purifi ed in a series of cross-fl ow fi ltration processes based on size exclusion.Suspended solids were removed from the reaction broth by cross-fl ow fl at sheet nylon prefi ltration membrane (pore size 100 μm).LMM peptides were purifi ed from the clarifi ed reaction broth by cross-fl ow fi ltration using tubular ceramic ultrafi ltration (UF) membrane (molecular mass cut-off (MMCO) 5 kDa).To diminish the concentration polarization near the prefi ltration membrane surface, constant transmembrane pressure (TMP) of 3•10 5 Pa was applied, whereas during LMM peptide purifi cation with UF membrane, diff erent TMPs and the use of a static turbulence promoter, a mechanical device promoting high shear on membrane surface, were considered.To obtain more purifi ed LMM peptides, 4-stage discontinuous diafi ltration with intermediate cleaning and volume concentration factor (VCF) of 2 were used.
Molecular mass of LMM peptides (M<5 kDa) in UF permeate was determined by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and liquid chromatography-electrospray ionization quadrupole time-of-fl ight mass spectrometry (LC-ESI-Q-TOF-MS).Effi ciency of the removal of allergenic proteins or peptide sequences was quantifi ed by an enzyme-linked immunosorbent assay (ELISA) method and characterized with LC-ESI-Q-TOF-MS-based bott om-up sequencing.The effects of purifi ed LMM peptides on Bacillus cereus and Pediococcus acidilactici HA6111-2 were also investigated together with the quantifi cation of antioxidant capacity that might be related to the promotion of bacterial growth.

Defatt ed soybean meal
Defatt ed soybean meal was kindly supplied by Pannon Vegetable Oil Production Ltd. (Foktő, Hungary).

Experimental setup
In this study, an in-house built membrane bioreactor, i.e. a bioreactor with a series of external membrane separation units, was used for low-molecular-mass (LMM) peptide production from defatt ed soybean meal.The schematic diagram of the experimental setup is shown in Fig. 1.
In detail, a 2.5-litre jar bioreactor (working volume 1 L) was used for the enzymatic digestion of soybean meal proteins.The bioreactor was equipped with a single pitched fl at-blade turbine.Two probes were placed into the bioreactor for measuring pH and temperature during the enzymatic reaction.The reactor had a water jacket and it was fi tt ed with a thermostat.A cross-fl ow fl at-plate membrane module (length 220 mm, width 120 mm and height 60 mm), made of stainless steel (SS316), was attached to the bioreactor externally.Three fl at-sheet nylon prefi ltration membranes were placed inside the module.The dimensions of a single prefi ltration membrane were: length 200 mm, width 100 mm, thickness 150 μm, active fi ltration area 1.4•10 -2 m 2 and pore size 100 μm.The permeate end of the prefi ltration membrane unit was attached to the temperature-controlled storage tank.Another cross-fl ow single-channel tubular membrane module, made of stainless steel (SS316), was fi tt ed with the storage tank.Inside that membrane module, a tubular ceramic ultrafi lter (UF) membrane with the molecular mass cut-off (MMCO) of 5 kDa (Membralox ® T1-70; Pall Corporation, Crailsheim, Germany) was placed.The active layer and support layer of the membrane were titanium oxide and aluminium oxide, respectively.The length, inner and outer diameters were 250, 7 and 10 mm, respectively, and active fi ltration area was 5•10 -3 m 2 .Transmembrane pressures (TMPs) of both types of membrane modules were monitored by pressure gauges and fl ow control valves, fi tt ed at the two opposite ends of the membrane modules.Feed fl ow rates were controlled using hydraulic diaphragm pumps (Hydra-Cell G03; Verder Hungary Kft ., Budapest, Hungary) and the inlet and outlet fl ow valves.A rotameter at the retentate end and a bypass valve were also used for controlling the fl ow rate in the UF membrane module.Stainless steel twisted tape static turbulence promoter was placed inside the ceramic membrane tube to create turbulence on the membrane surface.The promoter had an aspect ratio O tp =2, diameter d tp =6.5 mm, total length l tp =241 mm, pitch length l e =13.2 mm, number of mixing elements N TP =36 and thickness δ TP =1.2 mm.

Tryptic digestion of defatt ed soybean meal and membrane-based separation of LMM peptides
Prior to the experiment, defatt ed soybean meal powder with particle size of approx.200 μm was prepared in a laboratory ball mill (MM 400; Retsch Technology, Haan, Germany).Tryptic digestion was carried out using free (i.e.non-immobilized) trypsin in a bench-top bioreactor, operated in batch mode.Buff ered suspension of the soybean meal at appropriate concentration was prepared for enzymatic reaction and then the substrate was preincubated under fi xed operational temperature of 40 °C.Aft er preincubation, a fi xed concentration of enzyme (0.14 g/L) was added into the reaction medium to initiate the enzymatic hydrolysis.The reaction was performed under fi xed agitation speed of 175 rpm and programmed operating conditions up to 4 h.To optimize the enzymatic digestion, initial concentration of soybean meal (γsm 0 ), reaction temperature and pH were varied in the range of 25-100 g/L, 25-60 °C and 6.0-11.0,respectively.The pH of the reaction medium was maintained at 6.0, 7.0 and 8.0-11.0 with 50 mM of phosphate-citrate buff er, sodium phosphate buff er and Tris buff er, respectively.Isothermal conditions during enzymatic reaction were maintained with circulating water (warm/cold) through the jacket of the bioreactor.Aft er 4 h of enzymatic reaction, 800 mL of the reaction broth were pumped and pretreated with a cross-fl ow prefi ltration membrane separation process with retentate recirculation.Suspended solids from the reactor were drained properly aft er each experiment.The permeate of prefi ltration membrane was used for isolation of LMM peptides (M<5 kDa) by a tubular cross-fl ow UF membrane.In this case, LMM peptides were passed through the porous channel of the UF membrane as permeate, whereas trypsin (M=24 kDa) and residual high--molecular-mass peptides (M>5 kDa) were rejected.To reduce the concentration polarization near the prefi ltration membrane surface, constant operational TMP of 3•10 5 Pa was used.On the other hand, to optimize the purifi cation process of LMM peptides in UF membrane, diff erent TMPs, ranging from 2-5•10 5 Pa, and a static turbulence promoter were used.During fi ltration with UF membrane, constant retentate fl ow rate (Q R ) of 100 L/h was maintained.In order to obtain more purifi ed LMM peptides, four stages of discontinuous diafi ltration with intermediate cleaning and constant volume concentration factor (VCF) of 2 were adopted during separation in UF membrane.

Estimation of the degree of hydrolysis and kinetic constants of enzymatic hydrolysis of soybean meal protein
For each set of enzymatic hydrolyses, samples were withdrawn from the bioreactor through a capillary needle in 1-hour intervals throughout the whole hydrolysis period.Samples were evacuated into sample tubes and were immediately placed in a water bath at 90 °C to inactivate the enzyme.Temperature of the sample was reduced to 25 °C in a water bath (Precision COL 19; Thermo Scientifi c TM , Waltham, MA, USA), and 5 mL of 200 g/L of trichlorocacetic acid were added to the 5 mL of the sample.The mixture was left to stand in the water bath at 25 °C for 30 min to allow precipitation.Subsequently, the sample was centrifuged at 2800×g (Z206A; Hermle, Wehingen, Germany) at 20 °C for 30 min.If the pH of the supernatant was below 7.0, 1 mol/L of sodium hydroxide was used to neutralize it (pH=7.0).Protein concentrations in the sediment and supernatant were determined by the Kjeldahl method according to the ISO 20483:2013 (27) and the Lowry assay (28), respectively.Degree of hydrolysis (DH) (in %) of soybean meal protein was calculated according to the following equation: where γsp 1 is the concentration of soluble protein at any time in 10 % of trichlorocacetic acid (g/L), γsp 0 is the concentration of soluble protein at initial time in 10 % of trichlorocacetic acid (g/L) and γs 0 is the initial protein concentration in substrate (g/L) (29).
To estimate the maximum reaction velocity (v max ), inhibition constant (K s ) and the half-saturation constant (K m ), nonlinear regression analysis was performed with initial enzymatic hydrolysis reaction rate (r 0 ) and corresponding substrate concentration (γs): where k cat is turnover number (1/h) and γE is enzyme concentration (g/L) (30).
The Arrhenius equation (Eq. 3) was considered to estimate the apparent activation energy (E a ) of enzymatic reaction.Logarithm values of k cat vs. reciprocal of temperature diff erence were used to estimate the E a of enzymatic reaction.In the proposed investigation, reference temperature T*=298 K was considered: where k a is the constant of activation energy (1/h), R is ideal gas constant (kJ/(K•mol)) and T is reaction temperature (K) (31).

Estimation of thermodynamic constants of enzymatic reaction
In order to estimate the free energy change of the activation of enzymatic reaction (ΔG*), transition state theory was considered (32).The Eyring's absolute rate equation was adopted for this purpose: /4/ where h is Planck's constant (Js) and K B is Boltzmann constant (J/K) (31).
The activation enthalpy of enzymatic reaction (ΔH*) was estimated using the following equation ( 31 Permeate volumes (V) during the UF membrane fi ltration were collected at constant time (t) intervals.Permeate fl ux (J) of the UF membrane fi ltration was calculated according to the following equation: /9/ where A is eff ective membrane surface area (m 2 ) (33).
Change of permeate fl ux (ΔJ i ) (in %) in each discontinuous diafi ltration stage and TMP was calculated by the following equation during the UF membrane fi ltration: where J 0 is initial permeate fl ux (L/(m 2 •h)) and J f is fi nal permeate fl ux (L/(m 2 •h)).
Specifi c energy consumption of the UF membrane fi ltration Pressures at two opposite ends of the UF membrane module were recorded and from these, pressure drop (Δp) was calculated.Specifi c energy consumption (E s ) of the UF membrane fi ltration was calculated by the following equation ( 34):

Rejection of the enzyme by the membrane
The rejection of the enzyme by the UF membrane was calculated according to the following equation:

Estimation of water permeability through membrane and membrane hydraulic resistance
Before experiments, the membranes were compacted to prevent any possibility of changing the water permeability (L p ) through the membrane and membrane hydraulic resistance (R m ) during experimentation.Both types of membranes were compacted under the pressure of 5.5•10 5 Pa with Milli-Q water (Merck-Millipore) until the water fl ux became steady.The L p was estimated from the slope of water fl uxes versus corresponding operational TMPs (1-5•10 5 Pa).

/13/
In a similar way, R m was determined according to the Darcy's law for both types of membranes: /14/ where μ is dynamic viscosity of water (33).

Membrane cleaning
As the pore size of prefi ltration membrane was considerably large, there was low possibility of protein or peptide adsorption on the membrane surface.To clean off the polarization layer on the prefi ltration membrane surface, 10 g/L of sodium hydroxide solution was used at TMP of 5•10 5 Pa.Subsequently, water cleaning was done at the same operational TMP for 1 h.The UF membrane was cleaned thoroughly for 30 min with 10 g/L of Ultrasil P3-11 (Ecolab-Hygiene Kft ) and for 1 h with 10 g/L of citric acid with intermediate water cleaning steps aft er individual operation steps.Flow rate of 200 L/h without any TMP was used during the cleaning with Ultrasil and citric acid, while TMP of 5•10 5 Pa and a fl ow rate of 200 L/h were used during the cleaning with water.In each case, the water fl ux was found to regain its original value almost completely (i.e.value of water fl ux before experiment).

Total protein analysis
Total nitrogen as well as mass concentration of the protein in initial defatt ed soybean meal ( p ) were quantifi ed by the Kjeldahl method ( 27) with a semi-automatic Kjeldahl nitro gen analyzer equipped with a block digestion unit (Kjeldatherm ® ; Gerhardt, Königswinter, Germany).The  sp in the supernatant of the reaction medium treated with trifl uoroacetic acid and LMM peptides in membrane permeate were estimated by the Lowry assay method (28).Bovine serum albumin was used as a standard and colorimetric estimation was performed at the wavelength λ=393 nm with a UV-Vis spectrophotometer (Evolution TM 300; Thermo Scientifi c TM ).

Determination of tryptic activity
Activities of trypsin in the UF membrane permeate and retentate fractions were determined by a spectrophotometer (Evolution TM 300; Thermo Scientifi c TM ) using casein as substrate.Enzyme assay was performed with 50 mM of potassium phosphate buff er, pH=7.5.During enzymatic assay the incubation temperature of 30 °C was considered.Colorimetric determination was performed at the wavelength λ=660 nm.Activity of the enzyme was expressed in U/mL (35).

Free amino acid analysis
Concentrations of free amino acids (γFAA) in the reaction broth and UF membrane permeate were determined with an amino acid analyzer (AAA 400; INGOS s.r.o., Prague, Czech Republic), equipped with an Ionex Ostion LCP5020 ion-exchange column (200 mm×3.7 mm; Ionex SG LLC, Davis, CA, USA).To quantify the free amino acid content, 2 mL of liquid samples (hydrolysis broth and the UF membrane permeate) were mixed with 10 mL of 10 % (by volume) trichloroacetic acid and incubated for 1 h.Afterwards, the samples were fi ltered through disposable 0.45-μm polytetrafl uoroethylene (PTFE) membrane fi lters (Acrodisc ® syringe fi lters; Sigma-Aldrich, St. Louis, MO, USA).Ninhydrin (20 g/L in 2-methoxyethanol) at the fl ow rate of 0.25 mL/min was used for derivatization purposes in the amino acid analyzer.The instrument was equipped with a UV/Vis detector operating in the range of wavelength λ=440-570 nm.The column temperature was set to 55 °C, and lithium citrate buff er (pH=2.70-4.65)as mobile phase in a stepwise gradient was used at the fl ow rate of 0.3 mL/min.The pH gradient was set as follows: 0-24 min at pH=3.05, 25-49 min at pH=3.50, 50-74 min at pH=4.15, 75-137 min at pH=4.65 and fi nally, 138-171 min at pH= 2.70.

Molecular mass distribution of LMM soy peptides
Sodium dodecyl sulfate-polyacrylamide gel elec trophoresis (SDS-PAGE) was performed to estimate the molecular mass of the UF membrane permeate peptides, accord ing to the Laemmli method (36).A vertical gel electrophoresis cell (overall dimensions: width 0.14 m, length 0.22 m, height 0.19 m) together with a PowerPac™ high--current power supplier and a Gel-Doc EZ system (Bio--Rad, Hercules, CA, USA) were used in the experiment.The concentrations of running gel and stacking gel were 150 and 60 g/L, respectively.Standard protein marker (recombinant proteins, M=250-2 kDa; Precision Plus Pro- Activity of enzyme in retentate-

Activity of enzyme in permeate Rejection=
Activity of enzyme in retentate tein TM Dual Xtra, Bio-Rad) were used.Constant electrical potential diff erence of 200 V was applied during the experiment.
Furthermore, to determine the molecular mass distribution and sequences of the peptides present in UF membrane permeate, LC-ESI-Q-TOF-MS was used.Before LC--MS analysis, 3 mL of the UF membrane permeate were completed with 165 μL of acetonitrile (VWR International).The solution was vortexed, centrifuged for 10 min at 4100×g at room temperature (Z206A; Hermle).Solid phase extraction (SPE) was performed with SPE tube (SampliQ C18 Endcapped, 0.5 g, 3 mL; Agilent Technologies, Santa Clara, CA, USA).Prior to the sample loading, SPE tubes were conditioned with 5 mL of acetonitrile and equilibrated with 10 mL of 5:95 (by volume) acetonitrile/water solution.Aft er sample loading, tubes were washed with 5 mL of 5:95 acetonitrile/water solution and eluted with 1 mL of 80:20 (by volume) acetonitrile/water solution.The eluates were evaporated in a vacuum centrifuge (ScanSpeed Modulspin 32 (40), ScanVac; LaboGene, Lynge, Denmark).The residues were dissolved in 1 mL of 8:92 (by volume) acetonitrile/water solution containing 0.1 % (by volume) formic acid, fi ltered through 0.22-μm disposable PTFE fi lters (VWR International) and injected in the LC-ESI-Q-TOF--MS system.
For the LC-ESI-Q-TOF-MS experiments, a 6530 Accurate Mass LC-MS system (mass accuracy <2 ppm, mass resolution >10000; Agilent Technologies) was used.The ESI-Q-TOF-MS instrument was operated with an Agilent 6220 instrument with a dual ESI ion source in positive ionization mode.An Agilent 1200 high-performance liquid chromatography (HPLC) system, equipped with a Zorbax XDB C18 column (2.1 mm×50 mm×3.5 μm; Agilent Technologies) was used.HPLC mobile phases consisted of eluent A (0.1 % (by volume) formic acid in Milli--Q water; Merck-Millipore) and eluent B (acetonitrile with 0.1 % (by volume) formic acid).The fl ow rate was kept at 0.35 mL/min and the following gradient was used: 0-1 min 8 % B, 1-18 min up to 50 % B, 18-20 min up to 100 % B, 20-21 min 100 % B, 21-24 min down to 8 % B and 24-29 min 8 % B. Experiments with ESI-Q-TOF-MS were carried out at m/z=100-1700, while MS/MS experiments were accomplished at m/z=50-1700 with automatically fi xed collision energies for each peptide (slope 4 V/100 Da, off set 2 V) and with the following parameter set: precursor ion isolation in narrow MS/MS mode (m/z=1.3),detection frequency 4 GHz, fragmentor voltage 150 V, curtain voltage 65 V, fl ow rate of drying gas 13 L/min, capillary voltage 800 V, nebulizer pressure 2.75•10 5 Pa and gas temperature 300 °C.Prior to the experiments, the mass accuracy of the instrument was calibrated with ESI-L low concentration tuning mix (Agilent Technologies).MassHunter B.02.01 with SP3 and MassHunter Qualitative Analysis B.03.01 with SP3 soft ware (Agilent Technologies) were used for data analysis.
Molecular feature extraction soft ware was adopted to determine the molecular mass distribution and sequences of the peptides present in the UF membrane permeates.The initial set values of parameters for the Molecular Feature Extractor (part of MassHunter Qualitative Analysis soft ware) were as follows: target data type: small molecules (chromatographic), peak height: >10 000 counts, ion species: H + , Na + and K + , charge state: maximum charge of 5, isotope model: peptides, peak spacing tolerance: m/z= 0.0025, +7 ppm, compound fi lters: relative height ≥2.5 %, absolute height ≥5000 counts, compound ion threshold: two or more ions, retention time window: 0.3-14 min and m/z>200.Peak lists obtained from MS/MS spectra were identifi ed using OMSSA (open mass spectrometry search algorithm) v. 2.1.9(37) and X!Tandem version Sledgehammer (2013.09.01.1) (38).The search was conducted using SearchGUI v. 2.6.5 (39).Protein identifi cation was conducted against a concatenated target-decoy (40) version of the Glycine max (393 reviewed entries) and Bos taurus (1 reviewed entry for bovine cationic trypsin) complement of the UniProtKB (41) (394 (target) sequences).The decoy sequences were created by reversing the target sequences in SearchGUI.The identifi cation sett ings were as follows: trypsin with a maximum of two missed cleavages; 10 ppm as MS 1 and 0.02 Da as MS 2 tolerances; variable modifi cations: oxidation of M (+15.994915Da), acetylation of protein N-term (+42.010565Da), pyrroli done from E (-18.010565Da) and pyrrolidone from Q (-17.026549Da).Peptides and proteins were inferred from the spectrum identifi cation results using PeptideShaker v. 1.8.0 (42).Peptide spectrum matches (PSMs), peptides and proteins were validated at a 1 % false discovery rate (FDR), estimated using the decoy hit distribution.
ELISA method for the quantifi cation of soy allergens in powdered soybean meal and membrane permeate In order to reach the analytical range of 2.5-25 mg of soy protein per kg of the ELISA kit, sample dilution (up to 2000-fold) of the soybean meal powder and sample concentrating (up to 6.25-fold) for the UF membrane permeate were carried out according to the manufacturer's instructions.That is, for the dilution, 500 mg of sample were extracted with 1250 mL of extraction solution and the obtained extract was further diluted with the extraction solution to meet the analytical range.To acquire a concentrated solution, 5 mL of the UF membrane permeate were mixed with 20 mL of extraction solution.Colorimetric detection (A 630 nm ) was performed with a microwell plate reader (DIAReader ELx800G-PC; DIALAB GmbH, Wr.Neudorf, Austria).

Microbiological assay
Antibacterial activity of the LMM peptides from UF membrane permeate against Gram-positive Bacillus cereus (from the Culture Collection of the Department of Microbiology and Biotechnology [CCDMB], Szent István University, Budapest, Hungary) was investigated with the agar well method.A volume of 100 μL of 24-hour culture of B. cereus from liquid broth (6•10 6 colony-forming units (CFU)) was spread on the soybean casein digestive agar medium (Merck).Agar wells with the diameter of 5 mm were fi lled with 100 μL of the UF membrane permeate.Zone of inhibition was measured aft er 48 h of incubation at 37 °C.
The possible synergistic eff ect of the LMM peptides from UF membrane permeate on the growth of Pediococcus acidilactici HA6111-2 (from CCDMB) was investigated.Single microplate wells were fi lled with 135 μL of sterile soybean casein digestive medium (Merck), 135 μL of the UF membrane permeate and 30 μL of 24-hour inoculum of P. acidilactici HA6111-2 (6•10 6 CFU).Positive and negative controls were also considered to compare the results.Incubation temperature was set at 30 °C.Microbial growth was measured at the wavelength λ=595 nm by a microplate reader (Multiskan GO; Thermo Scientifi c TM ).Data were recorded at 30-minute intervals for the total microbial growth period of 24 h.

Antioxidant capacity
Antioxidant capacity of LMM peptides from UF membrane permeate was quantifi ed with the ferric reducing ability of plasma (FRAP) method using a UV-Vis spectrophotometer (Evolution TM 300; Thermo Scientifi c TM ).Ascorbic acid was used as reference and the absorbance was measured at 593 nm (43).

Statistical analysis
All experiments were performed in triplicate and the mean value with standard deviation (S.D.) was calculated.Statistical data analysis was conducted using a Microsoft Excel spread sheet (Microsoft Corporation, Redmond, WA, USA) and Minitab statistical soft ware v. 16 (Minitab Inc., State College, PA, USA).

Eff ect of initial concentration of soybean meal
To optimize the enzymatic hydrolysis, diff erent initial concentrations of soybean meal (γsm 0 ), ranging from 25-100 g/L, were used.Taking into account the total concentration of protein (γp) in the defatt ed soybean meal ((41.0±0.1)%, by mass), the substrate range covered approx.10-41 g of soybean meal protein per L. In Fig. 2a, soluble protein concentration (γsp) in the reaction broth was plott ed against time progress for diff erent γsm 0 .
It is observed that γsp increases with the increase of γsm 0 up to 75 g/L.Low substrate concentration (γs) in reaction medium increases the substrate solubility and reduces the intermolecular agglomeration of enzyme-substrate complex, which facilitates the enzymatic hydrolysis as well as product synthesis.Contradictorily, when γsm 0 >75 g/L, enzyme active sites are overloaded with the substrate molecules.A fraction of substrate binds with the active enzyme-substrate complex and produces inactive enzyme--substrate complex, which reduces the initial reaction rate (r 0 ), and the synthesis of soluble protein and low-molecular-mass (LMM) peptides (30,31).In Fig. 2b, diff erent r 0 are plott ed against corresponding γsm 0 .At limited γs (γsm 0 ≤75 g/L), the relationship between r 0 and γsm 0 is linear and aft er that it has a decreasing trend.The reasons for this observation have been described above.Accordingly, γsm 0 =75 g/L is considered optimum for the enzymatic hydrolysis.Aft er tryptic digestion of 75 g/L of soybean meal at 40 °C and pH=9, the concentration of free amino acids (γFAA) in reaction broth was 0.33 g/L.As trypsin is an endopeptidase and its activity is site-specifi c, the formation of free amino acids due to tryptic digestion was inherently low compared to the formation of soluble protein and LMM peptides.The γsp reached 8.4 g/L aft er 4 h of tryptic digestion of 75 g/L of soybean meal at 40 °C and pH=9.

Eff ect of temperature
Eff ects of temperature and pH on the enzymatic hydrolysis are shown in Fig. 3.The degree of hydrolysis (DH) (in %) of soybean meal protein increases with the increase of reaction temperature from 25 to 40 °C and at temperature exceeding 40 °C, it decreases (Fig. 3a).
For a typical enzyme-catalyzed reaction, temperature provides the activation energy, which increases the collision frequency between the enzyme and the substrate, as well as positively infl uences the turnover number (k cat ).At a particular temperature, substrate and enzyme come into proximity more frequently and the probability of enzyme and substrate binding increases, which promotes the product formation.With further increase of temperature, enzyme is denatured due to the breakdown of its structural peptide bond or change of its structural confi guration (30,31).In other words, substrate solubility increases with the increase of temperature, which facilitates the enzymatic hydrolysis.The r 0 along with respective reaction temperature is also shown in Fig. 3a.It is noted that r 0 is maximum at 45 °C and aft er that, it has a decreasing trend.The r 0 and DH (in %) were calculated based on the formation of soluble protein aft er 1 and 4 h of hydrolysis, respectively.The enzyme was active in the initial hydrolysis period (fi rst 1 h) and performed well at 45 °C, which is why the r 0 was high at that temperature.With time progress, enzyme was deactivated and consequently, the DH was reduced at 45 °C.Similar explanation was given in several cases (44,45).The value of maximum reaction velocity of enzymatic hydrolysis (v max ) was increased almost 2.6-fold by increasing the temperature from 25 to 40 °C.The values of kinetic constants and thermodynamic parameters of enzymatic hydrolysis are reported in Table 1.
The value of apparent activation energy of enzymatic reaction (E a ) was determined to be 54.87 kJ/mol.The change of the free energy (ΔG*) and enthalpy (ΔH*) of the activation of enzymatic reaction were calculated according to transition state theory (32).Positive values of ΔG* signify that the free energy change used for the activation of enzymatic reaction is not thermodynamically favoured.All ΔG* values (Table 1) are almost identical and their order of magnitude is expected for enzyme denaturation (46).Also, large positive values of ∆H* are characteristic for enzyme denaturation (32,(46)(47)(48).Negative values of activation entropy of enzymatic reaction (ΔS*) are also reported in Table 1.They signify that the activation of enzymatic reaction is supported by entropy and the thermal deactivation did not entail any signifi cant changes in the secondary and tertiary structure of the enzyme (49).Negative values of ΔS* are probably due to the compaction of the enzyme molecule and such changes can occur from the formation charge of neighbourhood molecules of enzyme or ordering of the solvent molecules (32,50).The parameter ΔH* provides the information about the number of broken non-covalent enzyme bonds and ΔS* provides the information about the net enzyme/solvent disorder change associated with the native structure of the enzyme to a transition state of enzyme confi guration (50,51).However, it is very diffi cult to estimate the number of broken non-covalent bonds required to form the transition state of the enzyme.Finally, the temperature of 40 °C is considered optimum for enzymatic hydrolysis.

Eff ect of pH
The catalytic activity of enzymes depends on the pH of the medium.The enzyme has ionic (functional) groups in its active site and due to variations of the pH of the medium, they are ionized in diff erent ways (51).In Fig. 3b, DH (in %) of soybean meal protein and r 0 at corresponding reaction pH are presented.It is observed that with increase of reaction pH, the DH (in %) and r 0 increase, with their maximum at pH=9.0 and 40 °C.The catalytic site of trypsin is characterized by the catalytic triad Asp102, His57 and Ser195 (52).At pH=9, the reaction medium is properly balanced with anionic and cationic ions (groups), which provide the stable ionization of the enzyme, its structural stability and formation of enzyme-substrate complex (30,31).At pH>9.0, the DH of soybean meal protein decreases.The deactivation process at high alkaline pH is possibly due to disulfi de exchange, which usually occurs at alkaline conditions.Similar explanation was given earlier during pectolytic reaction (53).Naidu and Panda (50) reported that a decrease in enthalpy and en-  tropy values occurs when pH of the reaction increases or decreases.According to them, enzyme folds during deactivation.At higher or lower pH values, the stable three--dimensional structure of enzyme is compressed, resulting in lower enzymatic activity and lower DH.According to the manufacturer (Sigma-Aldrich), the optimum pH for the proteolytic activity of trypsin is between 7.0 and 9.0.Also, signifi cantly high DH value (55-75 %) could be obtained during soy protein proteolysis in phosphate buff er due to the enzymatic and non-enzymatic deamidation of the glutamine in soy protein hydrolysate (54).In this investigation, it was found that DH increased 1.5-fold with changing the reaction pH=7 to pH=9.The down-regulated hydrolysis at pH=7.0 might be due to the presence of some residual trypsin inhibitors in soybean meal (55,56), which act in the presence of phosphate buff er at pH~7 (57,58).The optimum pH for the hydrolysis is considered to be pH=9.0.
Taking into account the total protein concentration (γp) in the actual experiment (approx.30.8 g of protein in 75 g/L of soybean meal), DH (in %) of soybean meal protein was found to be approx.28 % aft er 4 h of enzymatic hydrolysis at 40 °C and pH=9.0.Kong et al. (15) reported that DH was found to be 9 % when 50 g/L of isolated soy protein were treated with trypsin (enzyme/substrate=0.025) for 3 h at 55 °C and pH=7.0.Xie et al. (16) reported that DH was 22 % when 120 g/L of defatt ed soybean meal were treated with a combination of Protamex and trypsinase (enzyme/substrate=0.01).Enzymatic hydrolysis was performed at 56.78 °C and pH=6.10 for 10.72 h.In another case, DH was found to be 10.5 % when 30 g/L of isolated soy protein were hydrolyzed with trypsin (enzyme/sub-strate=0.025) at 37 °C and pH=7.5 for 4 h (20).As the DH of proteinaceous substance depends on the type of enzyme, the presence of peptide bonds in the substrate, hydrolysis time, reaction temperature and pH, and the analytical methods of estimating DH, the results are not directly comparable in all cases.

Separation of LMM peptides by membrane technology
Ultrafi ltration (UF) membrane was used for fi ne-tuned separation of LMM peptides from prefi ltration membrane permeate.A great emphasis was placed on optimization of the UF membrane separation.Aft er membrane compaction, R m and L p values of 5-kDa MMCO membrane were determined to be 3•10 13 m -1 and 3.3•10 -4 L/(m 2 •h•Pa), respectively.In all cases, more than 99 % correlation coeffi cients were achieved.
In a typical pressure-driven size-exclusion fi ltration, permeation is strongly infl uenced by the applied pressure on the membrane surface and diafi ltration stage.Permeate fl ux (J) curves of 5-kDa UF membrane, without and with a static turbulence promoter under a constant TMP (TMP=4•10 5 Pa) are shown in Figs.4a and b, respectively.
It is observed that in the 1st discontinuous diafi ltration stage, J declines rapidly, then gradually and eventually becomes steady with time.Similar trends at other TMPs were observed (results not shown).As the concentrations of unhydrolyzed proteins or peptides (M>5 kDa) in the storage tank were high, the concentration polarization, created by unhydrolyzed proteins or peptides on membrane surface, was more rapid in the 1st than in other discontinuous diafi ltration stages.The J increased with the increase of discontinuous diafi ltration stages because the fi ltration was performed in discontinuous diafi ltration mode with VCF of 2 and intermediate cleaning aft er every discontinuous diafi ltration stage.It was found that the application of the static turbulence promoter enhances the permeation in each discontinuous diafi ltration stage and it was almost similar in the 3rd and 4th discontinuous diafi ltration stages.Furthermore, it was also noted that the rate of J decline is low in fi ltration with UF membrane and static turbulence promoter.Static turbulence promoter increases the tangential velocity of fl uid across the membrane surface, which increases the turbulence and vorticity of fl uid near the membrane surface.As a result, the concentration polarization near the membrane surface is reduced and permeation fl ux through the membrane pores increases.Furthermore, static turbulence promoter provides the centrifugal force on the fl uid, which contributes to the driving force on the membrane surface.Due to the driving force in the presence of the static turbulence promoter, membrane gel layer resistance is reduced and J is increased.According to Figs. 4a and b, maximum J decline is in the fi rst stage of fi ltration.Therefore, the eff ect of TMP on permeation in the 1st discontinuous diafi ltration stage is shown in Fig. 4c, where the final permeate fl ux (J f ) and reduction of permeate fl ux (ΔJ i ) (in %) in the 1st discontinuous diafi ltration stage are plotted against corresponding operational TMP.It is observed that J f increases with the increase of operational TMP (TMP=2-4•10 5 Pa), and aft er TMP=4•10 5 Pa, there is no signifi cant change in the J f .With the increase of operational TMP, more driving force is applied on the membrane surface, which is the cause of low deposition of solute molecules near the membrane surface and higher permeation.At high operational TMPs (TMP=4•10 5 and 5•10 5 Pa), J reaches a constant value, designated as 'limiting fl ux' of membrane separation.This result might be the cause of: (i) saturation of concentration polarization layer near the membrane surface, as well as constant gel layer resistance (59), (ii) formation of a dynamic secondary membrane near the membrane surface due to high solvent permeation, and consequently thicker solute concentration on the membrane surface and point of incipient solute precipitation ( 60), (iii) osmotic pressure of the working solution is approximately equal to the applied pressure (61), and (iv) hydrodynamic resistance by boundary layer at high solute concentration on the membrane surface (62).In a similar way, the ΔJ i (in %) decreases with the increase of TMP and it becomes similar at TMP=4•10 5 and 5•10 5 Pa.In Fig. 4d, fi nal J f and ΔJ i (in %) in each discontinuous diafi ltration stage during the UF membrane separation are plott ed against the corresponding discontinuous diafi ltration stage, with TMP=4•10 5 Pa.It is observed that the J f increases with the increase of discontinuous diafi ltration stage.Consequently, the ΔJ i (in %) decreases with increase of the number of discontinuous diafi ltration stages and in the 3rd and 4th discontinuous diafi ltration stages, they become almost equal.Due to the dilution of the feed sample in storage tank before each discontinuous diafi ltration and intermediate cleaning of the membrane aft er each diafi ltration stage, J increases with the increase of the num-ber of discontinuous diafi ltration stages.The solute concentration in the storage tank was almost the same in the 3rd and 4th stages of the discontinuous diafi ltration, and because of that similar values of ΔJ i (in %) were obtained.In all cases, it is observed that static turbulence promoter has a positive infl uence on J.
Several researchers have reported about positive infl uence of the static turbulence promoter on membrane separation unit and fi ltration process.Krstić et al. (63) reported that the J improved more than 300 % and the membrane cleaning effi ciency also improved due to the incorporation of a static turbulence promoter, specifi cally with a Kenics static mixer during the microfi ltration of skimmed milk.Šereš et al. (64) made a comparison between the fi ltration with a static turbulence promoter and without such a device during the removal of non-sucrose compounds from sugar syrup.The highest J was achieved at 80 °C, with Q=400 L/h and TMP=10•10 5 Pa without the application of static turbulence promoter.During the fi ltration with static turbulence promoter, the highest J was achieved at TMP=6•10 5 -10•10 5 Pa and Q<100 L/h.J increased approx.65 and 30 % at 70 and 80 °C, respectively, when the static turbulence promoter was used.Moreover, J improvements were reported when Kenics static mixer was used in cross-fl ow microfi ltration of wheat starch suspension (65) and baker's yeast (Saccharomyces cerevisiae) separation (66).Gaspar et al. (67) investigated the effects of geometry of static turbulence promoter on oil-in--water microemulsion preparation.According to them, Kenics static mixer was a good choice for preparation of oil-in-water microemulsions.Positive infl uence of the static turbulence promoter on membrane emulsifi cation was also reported by Piacentini et al. (68).Ceramic membrane reactor with a static turbulence promoter was used for municipal wastewater reclamation.Due to the incorporation of the static turbulence promoter, J increased from 70 to 175 L/(m 2 •h) and average reduction rate of the chemical oxygen demand (COD) was more than 95 % (69).
Reduction of specifi c energy consumption (E s ) with high throughput of any equipment is one of the basic approaches to establishing the concept of 'process intensification'.In Fig. 5a, the eff ect of the presence or absence of static turbulence promoter in the UF membrane fi ltration at diff erent TMPs on E s is presented.The E s values in the UF membrane fi ltration with the static turbulence promoter are lower than in the fi ltration without this device.The reason behind this observation is that the static turbulence promoter provides the tangential velocity of fl uid across the membrane surface and the centrifugal force on the fl uid.At lower retentate fl ow rate (Q r ~100 L/h), the combination of these two eff ects, concentration polarization on membrane surface and membrane gel layer resistance are reduced, resulting in higher J with insignifi cant pressure drop at two opposite ends of the membrane (J >>∆p).The values of E s in the UF membrane fi ltration with static turbulence promoter are almost unchanged at operational TMP=3-5•10 5 Pa.In Fig. 5b, E s values are plott ed against diff erent discontinuous diafi ltration stages at constant operational TMP=4•10 5 Pa.It is observed that in all discontinuous diafi ltration stages, E s in the UF membrane with static turbulence promoter is lower than in the fi ltration without it.This can be explained by the fact that the fl uid velocity near the membrane surface increases due to the feed dilution in successive discontinuous diafi ltration stages and the high turbulence, provided by the static turbulence promoter.
The γLMM peptides in the UF membrane permeate was determined to characterize the separation characteristics of biomolecules in the UF membrane.In Fig. 6a, γLMM peptides in the permeate side of UF membrane is plott ed against corresponding operational TMP for the 1st discontinuous diafi ltration stage of fi ltration.It is observed that γLMM peptides in the permeate increases with the increase of operational TMP mostly up to 4•10 5 Pa, and it has similar value at TMP=5•10 5 Pa.Furthermore, it is observed that more LMM peptides are permeated when UF membrane fi ltration with static turbulence promoter is used.This may be justifi ed by the fact that due to high driving force and turbulence on the membrane surface caused by TMP and static turbulence promoter, there is low probability of adsorption of high-molecular-mass peptides (M>5 kDa) on membrane surface.High driving force provides higher convective fl ux.As the static turbulence promoter provides high shear force on the membrane surface, fl uid fl ows through the membrane channel at higher fl ow rate and solute dep-ositions on membrane surface are reduced.As a result, concentration polarization on the membrane surface is reduced and higher amount of LMM peptides passes through the porous membrane.In Fig. 6b, γLMM peptides in the UF membrane permeate are plott ed against the corresponding discontinuous diafi ltration stage at constant operational TMP=4•10 5 Pa.Most of the LMM peptides permeated in the fi rst discontinuous diafi ltration stage and the γLMM peptides in the permeate decreased with consecutive discontinuous stages due to the washing eff ect.Similar trends were observed in other cases also (results not shown).The purpose of using discontinuous diafi ltration in this fi ltration was to obtain more purifi ed LMM peptides.
The cumulative γLMM peptides increased with the increase of discontinuous diafi ltration stages.By mass balance, the cumulative γLMM peptides aft er the 1st, 2nd and 3rd stages of discontinuous diafi ltration were 3.35, 4.21 and 4.38 g respectively (original values of γLMM peptides were 8.33, 2.2 and 0.42 g/L in the permeate of the 1st, 2nd and 3rd stage of diafi ltration, respectively) when static turbulence promoter was used with UF at TMP=4•10 5 Pa.
Taking into account the total γp in the actual experiment (approx.30.8 g/L of protein per 75 g/L of soybean meal), approx.14 % (by mass) of total protein content could be recovered in the form of LMM peptides.The difference could arise from several sources.First, some proteins cannot be accessed for tryptic digestion because of either the irreversible denaturation caused by heat treat- ment during the soybean meal production or the lack of the use of chaotropic agents during digestion.Second, the mass of peptides containing intact disulfi de bridges may exceed the cut-off value of the used membrane.For example, the mass of one of the two disulfi de-bridged tryptic peptides of an abundant soy protein, glycinin G2 (Uni-Prot no.P04405; linked peptides: EQAQQNECQIQK + IE-SEGGFIETWNPNNKPFQCAGVALSR, where C indicates bridged cysteine residues) is over 4.4 kDa and clearly exceeds 5 kDa with even one tryptic miscut event.It was examined (in terms of enzymatic activity) that there was no contamination of enzyme in permeate.
Regarding all results shown in Figs. 5 and 6, it may be concluded that a combined strategy, i.e.TMP=3•10 5 Pa and three stages of discontinuous diafi ltration with VCF of 2 can be considered optimal for the separation using UF membrane with static turbulence promoter and an effective approach to establish the 'process intensifi cation' concept.Fig. 7 presents the mass concentrations of individual free amino acids (γFAA) in reaction broth and membrane permeates achieved at TMP 3•10 5 Pa and the 1st discontinuous diafi ltration stage.
Aft er enzymatic hydrolysis, the concentration of free amino acids in the reaction broth was 0.33 g/L, while in the permeate samples obtained with and without the static turbulence promoter it was 0.32 and 0.29 g/L, respectively.The partial loss of amino acids compared to their source (i.e.reaction broth) was possibly due to the resistance by the concentration polarization layer, created by non-hydrolyzed proteins or high-molecular-mass peptides (M>5 kDa).However, the amount of recovered free amino acids counts for a small part (about 1 %) of the initial total protein content, their composition carries important information as it is diff erent from the composition, achieved by the complete acidic digestion of soy protein (70,71).While lysine, arginine-derived ornithine, leucine and glutamic acid were the most abundant amino acids in both cases, the abundance of hydroxyproline, the amino acid found in soy cell wall proteins (72), is usually unreported in soybean meal studies (73).This might indicate the relatively higher accessibility of cell wall proteins in defatt ed soybean meal to tryptic digestion than of the enzymatically hydrolysable and extractable high abundant soy proteins in untreated soybean samples.

Characterization of LMM peptides
Molecular mass distribution of LMM peptides Permeate samples obtained by UF membrane fi ltration with and without static turbulence promoter were used to characterize and study the biological (antibacterial) activities of LMM peptides.According to the SDS--PAGE results, molecular mass of LMM peptides was smaller than 5 kDa.As porous (such as UF) membranes have asymmetric pore size distribution, there is a chance of contamination by untargeted high-molecular-mass peptides (M>5 kDa).Finally, no such contamination was detected in membrane permeate samples.Furthermore, molecular mass distribution of membrane permeate peptides was determined by LC-ESI-Q--TOF-MS analysis.Fig. 8 represents the distribution profi les of the peptides obtained in fi ltration with or without static turbulence promoter.
More than 96 % of the peptides (calculated as relative frequency) had molecular mass lower than 1.7 kDa and the highest molecular mass was found to be 3.1 kDa in both types of samples.These results prove that the 5-kDa MMCO UF membrane has a potential to reject the high--molecular-mass peptides (M>5 kDa).The molecular mass distributions of peptides obtained by the two diff erent processes are slightly diff erent, with lower molecular mass of the peptides in the fi ltrate obtained using static turbulence promoter.However, the diff erence cannot be considered signifi cant.Quantifi cation of soy allergens in the prepared LMM peptides Required sample preparation methods on which extraction effi ciency and selectivity of enzyme-linked immunosorbent assay (ELISA) depend were provided by the manufacturer.Therefore, absolute quantifi cation of allergenic proteins is challenging.However, relative quantifi cation eliminates the infl uence of operationally defi ned techniques on the determination process.Accordingly, the same ELISA kit was used in the soy allergen quantification of the soybean meal sample and of the soybean meal-derived LMM peptides.
The tryptic hydrolysis of soybean meal and the fi ltration process resulted in an almost complete removal of detectable soy allergenicity in the permeate containing the LMM peptides.In other words, >99.9 % of allergenic proteins were either removed by the membrane or hydrolyzed into non-allergenic peptides, calculated on mass basis (i.e.relative reduction based on defatt ed soybean meal powder).The application of the static turbulence promoter did not aff ect the fi nal elimination rate of allergenicity.Reduction of allergenicity >99.9 % could be achieved with and without this device.

Sequencing of LMM peptides
Peptide sequencing gives another important aspect of the study, especially in the viewpoint of decreased allergenicity.According to the data presented in Table 2, 12 soy proteins were unambiguously identifi ed with at least two residual peptides from the sample, using the static turbulence promoter, and 14 proteins were obtained without this device.Altogether, from the 16 proteins, 10 are considered allergenic, according to UniProt (41) and Gagnon et al. (74).On the other hand, as proteins are rejected by the membrane fi ltration, any residual allergenicity must come from the tryptic peptides because no allergenic soy protein is described in proteomic databases with M<5 kDa.Soybean meal proteins generally lose allergenicity aft er proteolytic digestion (75).Therefore, the sequencing information on LMM peptides in the permeate samples contributes to the background knowledge about hypoallergenicity.
In the detected allergenic soy proteins that have disulfi de bonds (indicated with lett er a in Table 2), no tryp-tic peptides could be found neither with the intact disulfi de bond nor with free cysteine residues.As neither unfolding agent, e.g.urea, nor reduction or alkylation were applied during tryptic digestion, the peptides with intact disulfi de bonds of the proteins identifi ed in the samples (see Table 2) have a theoretical molecular mass over 2.7 kDa, even without taking into account tryptic miscuts.That is, the bare tryptic hydrolysis assisted with the membrane fi ltration proved to be useful in decreasing the load of allergenic proteins or peptides in the fi nal samples.
Another remarkable diff erence between the two samples was the lack of most of the glycoprotein-derived peptides (indicated with lett er b in Table 2) as only one such allergenic protein was identifi ed in the sample obtained by the fi ltration with static turbulence promoter, while four were identifi ed in the sample obtained without the use of static turbulence promoter.This result might indicate higher retention of the glycoprotein-derived peptides by the membrane fi ltration if it is assisted with such an additional device.These characteristics are supported by the ELISA results, showing very low level of soy allergenicity.

Antibacterial activity
In our study, it was observed that the prepared LMM peptides have antibacterial activity against one of the common food pathogen bacilli, Bacillus cereus.Fig. 9a shows the inhibition zones with diameters of 15 and 10 mm obtained by UF membrane fi ltration with and without static turbulence promoter, respectively.It is reported that several plant-based peptides disrupt bacterial cell membranes, as they can create stable pores through which the cellular content leaks out (76).Furthermore, Matsuzaki (77) reported that antibacterial mechanism of plant-based (such as soybean) peptides is detergent-like.The Gram-positive bacteria have negatively charged membrane phospholipids, i.e. teichoic acid and lipoteicho- ic acids in their peptidoglycans, which presumably facilitate the electrostatic interactions with positively charged LMM peptides from soybean.Similarly, Wieprecht et al. (78) reported that cationic or amphiphatic hydrophobic peptides interact with the negatively charged bacterial cell membrane, which is the cause of membrane thinning or micellization.Patrzykat and Douglas (79) found that the extent of binding ability of peptides with bacterial cell membrane lipopolysaccharides is neither directly nor inversely proportional to the activity of peptides.The sequence, peptide concentration and chemical composition of the bacterial cell membrane infl uence the mode of antibacterial action of peptides.
Fig. 9b shows the growth curves of Pediococcus acidilactici in membrane permeate with LMM peptides.Interestingly, it was found that the purifi ed LMM peptides could promote the growth of benefi cial microbe P. acidilactici, which produces H 2 O 2 during aerobic fermentation, with a fatal eff ect on its growth.Peptides with hydrophobic amino acid residues in the membrane permeate presumably donate an electron to H 2 O 2 , and as a result, neutralize the reactive oxygen species and promote the growth of P. acidilactici.Indeed, Takeuchi (80) reported that special fractions of defatt ed soybean fl our are growth-promoting factors for Pediococcus sp.Herawati and Ishizaki (81,82) also reported that soybean hydrolyzate 'Mieki' enhanced the growth of Pediococcus sp.Furthermore, it is well documented that Soytone (enzymatic hydrolyzate of soybean; Difco, Detroit, MI, USA) enhances the growth of Pediococcus sp.(83,84).
It is well documented that plant-based LMM peptides have considerable antioxidant capacity and that they can neutralize the eff ect of H 2 O 2 (85).The antioxidant capacity of LMM peptides (expressed in ascorbic acid equivalents) obtained with fi ltration using static turbulence promoter was 60 mg/L, which equals 13.7 mg/g of LMM peptide from soybean meal.This result is quite high compared with the tryptic digestion of palm kernel cake protein, which has an antioxidant capacity of 11.4 mg/g of peptide (86).Even though the antioxidant capacities of diff erent matrices obtained by diff erent analytical methods cannot be directly compared, the antioxidant capacity of the LMM soybean meal peptides could possibly explain why the growth of P. acidilactici was promoted, thus highlighting an additional bioactive property of these peptides.

Conclusion
Valorization of by-products from food and feed industries has grabbed considerable att ention.Without any contradiction, production of bioactive peptides from a by--product of vegetable oil processing industry, in this case defatt ed soybean meal, is one of the promising approaches on the platform of waste valorization.In the present investigation, production of hypoallergenic antimicrobial low-molecular-mass (LMM) peptides from defatt ed soybean meal was carried out by the combination of enzymatic treatment and membrane fi ltration based on size exclusion to provide the information for scale-up of the process in food and biochemical process industries.Our approach addresses the production of hypoallergenic antibacterial LMM peptides from defatt ed soybean meal instead of using soy protein isolate, which is the fi rst such att empt to the best of our knowledge.The proposed method is expected to be a progressive step in the fi eld of recycling and reuse of agricultural waste materials.

Fig. 1 .
Fig. 1.Schematic diagram of the experimental setup.TIC=temperature indicator controller, p 1 and p 2 are pressure gauges

): / 5 / 6 / 8 /
The activation entropy of enzymatic reaction (ΔS*) was estimated according to the following equation (31): /The free energy changes of substrate-enzyme binding (∆G b ) were estimated according to the following equation (31): /7/ where γ0 is standard concentration of 1 g/L (assumption value).The enzyme-substrate binding enthalpy (∆H b ) and enzyme-substrate binding entropy (∆S b ) were calculated by regression analysis (31): /Estimation of membrane permeate fl ux

Fig. 2 .
Fig. 2. Changes in: a) mass concentrations of soluble protein in the reaction broth with time, and b) initial enzymatic hydrolysis reaction rate (r 0 ) of tryptic digestion, using diff erent initial mass concentrations of soybean meal (γsm 0 ) at reaction temperature of 40 °C and pH=9.0

Fig. 3 .
Fig. 3. Degree of hydrolysis (DH) and initial enzymatic hydrolysis reaction rate (r 0 ) aft er 4 h of hydrolysis at: a) diff erent reaction temperatures with initial concentration of soybean meal (γsm 0 ) of 75 g/L and reaction pH=9.0, and b) diff erent reaction pH values with initial concentration of soybean meal (γsm 0 ) of 75 g/L and reaction temperature of 40 °C

Fig. 4 .
Fig. 4. Permeate fl ux (J) at transmembrane pressure (TMP) of 4•10 5 Pa with discontinuous diafi ltration: a) without and b) with a static turbulence promoter during ultrafi ltration (UF), c) fi nal permeate fl ux (J f ) and change of permeate fl ux (ΔJ i in %) at the 1st discontinuous diafi ltration stage at diff erent TMPs during UF, and d) fi nal permeate fl ux (J f ) and change of permeate fl ux (ΔJ i in %) at diff erent discontinuous diafi ltration stages at TMP=4•10 5 Pa during UF

Fig. 9 .
Fig. 9. Zone of inhibition by the low-molecular-mass (LMM) peptides in ultrafi ltration membrane permeate against Bacillus cereus (a), and growth of Pediococcus acidilactici HA6111-2 in the presence of LMM peptides from membrane permeate (b)

Table 1 .
Reaction kinetic and thermodynamic parameters of the enzymatic hydrolysis cat =turnover number of the enzymatic hydrolysis reaction, K m =half-saturation constant of the enzymatic hydrolysis reaction, ∆G*=free energy change of the activation of the enzymatic reaction, ∆H*=activation enthalpy of the enzymatic reaction, ∆S*=activation entropy of the enzymatic reaction, ∆G b =free energy changes of the substrate-enzyme binding, ∆S b =enzyme-substrate binding entropy, ∆H b =enzyme-substrate binding enthalpy