Membrane fractionation of herring marinade for separation and recoveryof fats, proteins, amino acids, salt, acetic acid and water
Lene Fjerbæk Søtoft
, Juncal Martin Lizarazu, Behnaz Razi Parjikolaei, Henrik Karring, Knud V. Christensen
University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark
a r t i c l e i n f o
 Article history:
Received 14 November 2014Received in revised form 11 February 2015Accepted 23 February 2015Available online 2 March 2015
Fish food wasteHerring brineProtein characterizationMembrane filtrationFractionation
a b s t r a c t
In the production of marinated herring, nearly one ton of acidic saline marinade is produced per 1.5 tonsherring fillet. This spent marinade contains highly valuable compounds such as proteins and amino acids.Membranes are suited to recover these substances. In this work, six membrane stages are employed:microfiltration (MF) (0.2
m), ultrafiltration (UF) (50, 20, 10 and 1 kDa) and nanofiltration (NF).The most promising stages are 50 kDa UF and NF based on SDS–PAGE analyses and total amino acidconcentration. The 50 kDa stage produces a protein concentrate (>17 kDa). NF produces a retentate con-taining sugars, amino acids and smaller peptides and a NF permeate containing salt and acetic acid readyfor reuse. 42% of the spent marinade is recovered to substitute fresh water and chemicals. The wastewater amount is reduced 62.5%. Proteins are concentrated 30 times, while amino acids and smaller pep-tides are concentrated 11 times.
 2015 Elsevier Ltd. All rights reserved.
1. Introduction
The seafood industry for human foods is a very water-intensiveindustry (Afonso and Borquez, 2002; Almas, 1985; Matthiassonand Sivik, 1978). The waste water is generally characterized by ahigh organic load and a varying salt content (Vandanjon et al.,2002). At the same time, there is a huge potential for recovery of valuable compounds of marine srcin and make-up water fromthe waste fractions. This is important for the (1) better utilizationof valuable marine compounds and new value-added by-products,(2) reduction in raw material consumption for improved processeconomy and (3) reduction of the environmental impact of foodproduction.A technology which can reduce water consumption in water-intensive industries is membrane separation (Afonso andBorquez, 2002; Almas, 1985; Matthiasson and Sivik, 1978).Additionally, membrane technology can offer separation/recoveryof particles and molecules in very specific ranges making it veryinteresting for byproduct separation.Microfiltration (MF), ultrafiltration (UF), nanofiltration (NF) andelectrodialysis are already seen as established technologies as dis-cussed by Galanakis (2012), but are sensitive to fouling due to the nature of the raw material (Galanakis, 2012). Several studies havereported the use of MF, UF, NF and reverse osmosis (RO) forseparation, removal or recovery of organic material from fishindustry waste water streams (Dumay et al., 2008; Ferjani et al.,2005; Matthiasson and Sivik, 1978; Li et al., 2006, 2008; Perez-Galvez et al., 2011; Stine et al., 2012; Vandanjon et al., 2002, 2009).Because membrane processes are carried out at a relative lowtemperature, they offer an improved preservation of the concen-trated compounds such as proteins compared to traditional ther-mal or chemical processes (Dumay et al., 2008). Matthiasson and Sivik (1978) were the first to demonstrate the usefulness of UFand RO for processing various waste waters from herring process-ing including spent herring marinade with 15–22 wt% salt. The aimof that study was to recover a protein concentrate and reduce theorganic load in the waste water. The waste water COD (ChemicalOxygen Demand) load was reduced by up to 97% and proteinwas concentrated up to 10 wt% (Matthiasson and Sivik, 1978).These authors suggested combining membrane filtration withevaporation. Membrane filtration would then remove 65–90% of the water and make a 15–20 wt% commercial protein concentrate,which then by evaporation could be processed into 30–40 wt% pro-tein (Matthiasson and Sivik, 1978).Ceramic NF membranes (1 kDa) were used by Afonso andBorquez (2003) to concentrate protein from fish meal waste water.Aneconomicalassessmentofa10 m
/hfishmealwastewatertreat-ment plant showed a rate of return of 17% and a feasible process(Afonsoetal.,2004).Sincethen,therehasbeenahugedevelopmentwithin membrane processes, an increased interest in value-addedmarine products and a rise in environmental awareness.
 2015 Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +45 6550 7479.
E-mail address:
Contents lists available at ScienceDirect
 Journal of Food Engineering
Itisexpectedthatmembranefractionationofe.g.proteinhydro-lysatesisnotcharacterizedbyasharpcut-off.Bourseauetal.(2009)observed this in the fractions obtained after a UF (4000 Da cut-off)and NF (300 Da cut-off) separation sequence of two fish proteinhydrolysates. Size-exclusion chromatography showed how largermoleculeswereretainedintheretentateandthesmallermoleculeswere present in the permeate, but with a mixed retention in theintermediate molecular weight (MW) region (Bourseau et al.,2009).Beaulieuetal.(2009)fractionatedaherringhydrolysatewith similar results. A 50 kDa retentate had the highest total amino acidcontent (74% dry matter), while RO retentate contained mostminerals (42% dry weight). The used stages were 0.3
m, 50 kDa,10 kDa, 1 kDa, 200 Da (NF) and <200 Da (RO).The observed MW cut-off for a specific process is a combinationof the chemical and filtration properties of the membrane, sec-ondary layer(s) and feed, which can cause differences betweennominal and observed cut-offs. Rejection of bovine serum albuminby a 150 kDa ceramic membrane has been found to depend on pHand whether or not 10 wt% NaCl is in the model solution. With10 wt% NaCl and pH 9.0, the lowest rejection of 96.5% is obtainedwhile at lower pH (4.8 and 6.8) rejection is >99% also when NaClis present (Kuca and Szaniawska, 2009).The presence of salts can reduce the value of protein or aminoacids concentrate, but diafiltration can be used to purify the con-centrates even more. Diafiltration is a membrane process wherenew solvent is added to an existing concentrate. The smallerpermeating molecules such as sugars or salts are then washedaway during a filtration as permeate, while the largest moleculesfor instance proteins are retained and purified by the membrane.This can be carried out as a continuous or batch process and canbe a way to remove smaller molecules from a retentate, when ahigher concentration of the largest retained molecules is of interestas done by Taheri et al. (2014) on herring brine.As organisms of marine srcin are adapted to an environmentvery different from the terrestrial, they produce numerous inter-esting compounds such as pigments, proteins, polysaccharidesand lipids. Additionally, the market and interest in marine nutra-ceuticals are growing significantly (Rasmussen and Morrissey,2007). Examples could be marine proteases (Bougatef, 2014) and other bioactive peptides (Picot et al., 2010).The main aim of the present study is to recover valuable frac-tions from a fish industry waste and characterize the propertiesand potential use of each fraction. Additionally, the purpose isto reduce the amounts of saline waste water with high organiccontent discharged from the plant. This work is unique in thenumber of product fractions (six consecutive membrane stages)and the scale of the experiments with a starting volume of 120 L of spent herring marinade. The multiple objectives of thiswork can be specified as (1) recovery of organic particles, (2)recovery of proteins and amino acids, (3) recovery of fats, (4)recovery of sugars, (5) recovery of water and chemicals of suffi-cient quality for reuse and (6) volume reduction of waste for fur-ther treatment.
2. Material and methods
 2.1. Materials
A fresh sample of spent herring marinade (
120 L) was pro-vided by the Danish fish food producer LAUNIS FiskekonservesA/S. The sample was kept at 3–7
C. Before addition of the herring,the marinade is composed of water, acetic acid and salt. Thus, fishmeat residues, fat, proteins, peptides and free amino acids can beexpected to be present in the marinade in addition to acetic acidand salt after marinating the herring under anaerobic conditionsand subsequent removal of the fillets.An inspection of the marinade showed visible fish meat resi-dues and a fat layer. This had to be removed prior to any filtrationto protect the membranes. Due to the low temperature, the fatsolidifies on top of the liquid and can be removed efficiently byskimming (no visual oil droplets left).Chemicals (acetonitrile: HiPerSolv Chromanorm from VWR,PROLAB and water) for HPLC were of UV-grade, whereas cleaningchemicals for the membrane setups (citric acid and NaOH, VWR)were of food grade quality. Water used for membrane cleaningwas ion-exchange quality (conductivity below 10
S cm
 2.2. Methods 2.2.1. Sieving and membrane filtrations
A series of membrane filtrations with various pore sizes hasbeen carried out. Sieving is used as pretreatment and has beendone as manual batch sieving with dead-end sieves. Their charac-teristics can be seen in Table 1.The cross flow membrane unit used is a Labstack M20 for flat-sheet membranes (Alfa Laval). The retentate is recycled to the feedtank in order to concentrate the retentate. The setup is equippedwith a feed pump (Hydracell), an inline feed heat exchanger, twofeed side pressure gauges and a pressure control valve on theretentate side. Feed flow rate was controlled by adjustment of the pump speed. Permeate is collected separately during filtra-tions. A weight (Dansk Vægt Industry A/S, 0–35 kg) was used tomeasure the permeate flow rate. Temperature was controlled byconnecting a cooling unit (Heto HMT 200) to the heat exchanger.Additionally, the feed tank was cooled by insertion into a coolingtank during NF to reduce the temperature increase during opera-tion (Fig. 1). During each run, only one type of membrane was usedat a time. The fractions and samples are cooled immediately at 3–7
C after a filtration. The membranes used in each run can be seenin Table 2 specified with either pore size, molecular weight cut-off (MWCO) or rejection.The volume reduction (VR) for a specific fraction is calculatedas:
 Concentrate volume
Initial volume
 2.2.2. Dry matter and ash content 
Dry matter (DM) and ash content measurements were per-formed at 105
C and 550
C, respectively. The analyses were donein triplicate according to DS 204:1980.
 2.2.3. Protein quantification by absorbance at 280 nm
Total protein concentration in the different marinade fractionswas determined by measuring the absorbance at 280 nm (Beavenand Holiday, 1952; Layne, 1957). For unknown complex proteinmixtures, it is commonly accepted that 1 absorbance unit at280 nm equals 1 mg/mL protein (light path length of 1 cm). Thefractions were diluted in pure marinade solution (9.0 wt% NaCl,2.0 wt% acidic acid, pH 4.15) to obtain absorbance values in therange 0.1–0.7. Subsequently, absorbance at 280 nm (A280) wasmeasured for 1 mL samples using the pure marinade solution asreference. Dilutions and A280 measurements were performed intriplicates. Finally, the measurements were corrected for the dilu-tion factor to obtain absolute A280 values for the fractions.
 Table 1
Sieves used for pretreatment of the herring marinade.
Sieve area (mm) Material Supplier Mesh size (mm)200
50 Stainless steel Retsch 0.5200
25 Stainless steel Retsch 0.18201
25 Stainless steel Retsch 0.145202
25 Stainless steel Retsch 0.04540
 L. Fjerbæk Søtoft et al./Journal of Food Engineering 158 (2015) 39–47 
 2.2.4. Sodium dodecyl sulfate–polyacrylamide gel electrophoresis(SDS–PAGE)
The protein composition of the herring marinade fractions fromthe different sieving and filtration steps were analyzed by SDS–PAGE. Direct electrophoretic analyses of the raw fractions failedprobably due to the high concentrations of salt and lipids in thesamples. Therefore, the polypeptides were prepared for SDSPAGE by methanol–chloroform precipitation according to Wesseland Flügge (1984). After precipitation of the proteins the air-driedpelletsweredissolvedinSDS–PAGEsamplebuffercontainingafinalconcentration of 5 mM dithiothreitol. SDS–PAGE was performedusing hand-cast 4–16% gradient gels (7.3 cm
8.3 cm
1 mm)and the SDS–PAGE Buffer System (Laemmli, 1970). The separatedproteins were visualized by Coomassie Brilliant Blue staining.Relative quantification of three distinct and abundant proteinbands migrating at about 17 kDa (band a), 47 kDa (band b), and64 kDa (band c) were achieved by densitometry using AdobePhotoshop CS5.1 software (version 12.1
32) and according to Xuet al. (2010). The staining intensities were normalized in relationto the 28.5 kDa molecular marker band. Methanol–chloroform pre-cipitation, separation by SDS–PAGE, and quantification by densito-metry were performed in triplicates.
 2.2.5. High performance liquid chromatography (HPLC)
Qualitative analyses for ethanol, glycerol, fructose, glucose, lac-tose and sucrose were done by isocratic HPLC (Agilent series 1100)using a Luna 5
m NH2 100 Å 250
4.60 mm column with 65:35(vol%) acetonitrile:H
O as eluent. The column temperature was25
C, injected sample size 5
L, the eluent flow 0.5 mL/min andthe analysis time was 20 min. The analyses were done in duplicate.
3. Results and discussion
 3.1. Initial characterization of herring marinade
The initial composition of the marinade prior to addition of fishfillets is 9.0 ± 1.0 wt% NaCl, 2.0 ± 0.3 wt% acetic acid and a pH of 4.0–4.3. The herring marinade has additionally been characterizedin order to plan the membrane fractionations accordingly. The ini-tial chemical composition and characteristics of the marinade canbe seen in Table 3 and Fig. 2. An initial SDS–PAGE analysis of the untreated herring marinadehas been performed to characterize the sizes of the proteins pre-sent. The results can be seen in Fig. 2. It shows a few predominantbands in the range of 8–20 kDa and the presence of some proteinswith a molecular weight of 
25 kDa. Furthermore, several proteinsof various sizes between 30 kDa and
150 kDa are present in themarinade. Most of these bands are likely the result of proteolytic
MembraneHXPumpFeed tankPermeate tankValve
Fig. 1.
 Experimental setup (HX: heat exchanger,
: temperature, P1 and P2: retentate pressure gauges).
 Table 2
Flat sheet membranes used for herring marinade fractionation.
Process MembranetypeMaterial Pore size/MWCO/rejectionMF Alfa Laval-MFP2Fluoro polymer 0.2
mUF50 DSS-GR51PP Polysulphone 50 kDaUF20 Alfa Laval-FS61PPFluoro polymer 20 kDaUF10 DSS-ETNA10PPComposite fluoropolymer10 kDaUF01 Alfa Laval-ETNA01PPFluoro polymer 1 kDaNF Alfa Laval-NF Polyamide onpolyesterReject >98% MgSO
(2000 ppm, 9 bar, 25
 Table 3
Initial chemical composition and characteristics of herring marinade.
Analysis ValueDry matter 10.6 wt%Protein
3.9 wt%Fat 0.22 wt%pH 4.36
Measured with Dumas method for nitrogen determination.
Fig. 2.
 SDS–PAGE of the herring marinade. Left lane, the molecular weight markerwhich indicates masses in kDa. Right lane, herring marinade sample. Arrowsindicate the nominal MWCOs for the membranes used in the study.
L. Fjerbæk Søtoft et al./Journal of Food Engineering 158 (2015) 39–4
degradation of larger protein fragments as shown by Andersenet al. (2007). They observed significant proteolytic activity bothin the herring and brine during the ripening process leading tofragmentation of major proteins such as actin (42 kDa) and myosin(200 kDa), appearance of smaller proteins and peptides, andsmeared bands indicating heterogeneous proteolytic events.Based on the initial characterization of the marinade and therange of available commercial membranes a plan for the sequenceof membrane filtrations and subsequent characterization of thefractions were developed (Fig. 3). These fractionations serve as abasis to identify an optimized number of fractions for further eco-nomic and qualitative analysis.The initial aim is to separate valuable fractions of fats, fish meatand organic particles, proteins, peptides and amino acids as well aschemicalsand water of sufficient qualityfor reuse combinedwith avolume reduction of waste for further treatment.
 3.2. Fractionation of herring marinade
A sequence of membranes with different MWCOs has been usedto get a broad array of fractions with different molecular sizeranges for later characterization.As pretreatment to reduce membrane fouling, solidified fat isremoved upon cooling and then the marinade is sieved to removevisible fish meat particles. The sieving retained meat and proteinson the sieves (visual inspection). Even though the marinade is stillturbid after the sieving, the marinade is processed by MF as the ini-tial membrane filtration. Several UF MWCOs are chosen in order toinvestigate the characteristics of the individual fractions. This isspecifically interesting for the protein/peptide content as it has avery broad variation in molecular mass compared to e.g. freeamino acids. The permeate from the higher MWCO is then usedas feed for the following MWCO. Process characteristics can beseen in Table 4.
 3.2.1. Microfiltration
MF has been carried out as a constant pressure filtration with atransmembrane pressure (TMP) of 
0.9 bar at room temperature.Retentate is returned to the feed tank to further concentrate theretained fraction. The final concentrate and permeate are thenundergoing subsequent analysis. The permeate is feed for the fol-lowing UF50 (Fig. 3). The results of the analyses are covered inSection 3.3.An intermediate cleaning is performed with NaOH at pH 9,before new feed was added to complete the treatment of all themarinade. Clean water flux data before and after the cleaning showthat the membrane’s clean water flux of initial 190 kg/(m
h) at0.9 bar decreases to 60 kg/(m
h) after MF and is restored to145 kg/(m
h) after the cleaning. This means that 75% of the initialflux is restored.
 3.2.2. Ultrafiltration
The permeate generated in the MF step is then fractionatedwith UF. Four UF MWCOs have been used in series: 50 kDa,20 kDa, 10 kDa, and 1 kDa. The overall observation for UF is a fastinitial flux decline followed by a more steady performance for thelater part of the filtration.Due to the high number of stages, the feed for the UF processescan be considered as efficiently pre-treated, but a high initial fluxdecline is still observed for herring marinade filtrations. This isespecially severe for UF50 and UF01, while not as severe forUF10 and UF20. This is caused by the higher amount of retainedmaterials during UF50 and UF01 (see Section 3.3) compared toUF20 and UF10. A 0.2
m MF pretreatment and subsequent UF10filtration of herring marinade showed a large flux decline in theUF10 stage and a 10 times lower flux (Lyder, 2014) than obtainedin this work where intermediate UF50 was done. An intermediatestage of UF50 is therefore an option to increase the UF10 flux.The fast initial decline can be due to accumulation of moleculesat the membrane surface, which increases local osmotic pressureand/or builds a fouling layer and hence reduces the effective driv-ing force. Clean water flux measurements after the filtrations show
83% recovery of the initial flux, so the effect is mainly reversible.UF10 showed a higher initial clean water flux recovery comparedto the recovery of UF01, while the lowest clean water flux recov-eries were seen for UF20 and UF50. This might be due to
Sugarsamino acidsPeptidesProteins
Fish meat
Saline water for reuse
Herring marinade
Fat fraction
MF Major particulatesUF20ProteinsUF10 ProteinsUF01 Proteins
Fig. 3.
 Fractionation sequence. The number following UF indicates MWCO in kDa,where i.e. UF50 is UF with 50 kDa MWCO.
 Table 4
Process characteristics during filtrations of herring marinade fractions.
Process Initial flux(kg/(m
h))End flux(kg/(m
h))Crossflow(kg/min)TMP (bar) Filtrationtime (min)MF 48 20 5 0.9 330UF50 95 30 5 7.8 215UF20 250 150 5.56.6 5.96.0 45UF10 120 80 6.5 5.46.3 70UF01 35 5 6.5 4.26.7 410NF 16 4 6.5 26.429.1 45042
 L. Fjerbæk Søtoft et al./Journal of Food Engineering 158 (2015) 39–47 
of 9