Salt reduction in film‐ripened, semihard Edam cheese

As a potential measure to improve public health, this study aimed to reduce the sodium (Na) content of film-ripened, semihard Edam cheese to ≤0.4 g Na/100 g (≤1 g NaCl/100 g), while retaining typical quality and safety characteristics. For this, mineral salt substitutions containing potassium (K) were compared with simple NaCl reduction in brine, alongside an adjustment of starter cultures in an effort to enhance taste. Desired Na and K values were achieved, and microbial quality was not compromised in Na-reduced Edam after six weeks of ripening. However, all Na-reduced cheeses tasted bitter and were therefore organoleptically unsatisfactory.


INTRODUCTION
High intake of the mineral sodium (Na), which forms an ionic compound with chloride (NaCl), can increase blood pressure and is therefore an important risk factor for cardiovascular and kidney disease. While the World Health Organization (WHO 2012) recommends ≤5 g NaCl as the daily sodium intake for adults, the German Nutrition Society (DGE) recommends ≤6 g NaCl/day (Strohm et al. 2016). However, in Germany 70% of women and 80% of men consume more salt than recommended. Indeed, an intake of >10 g NaCl/day is reached by 39% of German women and 50% of German men (Strohm et al. 2016). Therefore, the German Federal Ministry of Food and Agriculture promotes the 'reformulation' of prepacked food with less salt, but also with less fat and sugar, and has initiated a comprehensive innovation research programme to support the national reformulation strategy (Federal Ministry of Food and Agriculture 2018).
Dairy products, including cheese, contribute to 10-11% of NaCl consumption in the nutrition of German women and men (Max Rubner-Institut 2008). Improvements in cheese technology, as well as hygienic progress, theoretically allow for a reduction in the salt content. However, a typical semihard cheese such as Edam still contains about 0.6-0.8 g Na/100 g (1.5-2.0 g NaCl/100 g) in Germany.
NaCl has a complex function in the manufacturing process of cheese and affects the microbiological, physicochemical and biochemical, rheological and sensory properties (International Dairy Federation 2014). Therefore, a reduction of NaCl in cheese has to consider many different aspects impacting cheesemaking and product characteristics. The effects of sodium reduction on functionality, sensory properties and public health were reviewed by Cruz et al. (2011), emphasising the need for further knowledge regarding acceptable salt levels in Na-reduced cheese.
Much scientific work focusing on the reduction or substitution of NaCl in ripened cheese was published during the last three decades. The aim of studies carried out by Barth et al. (1989) and Prokopek et al. (1990) was to manufacture an organoleptically acceptable, naturally ripened Edam cheese with ≤0.45 g Na/100 g. The resulting cheeses met the requirements for a standard variety during a shelf life of 12 weeks. However, partial substitution of Na with potassium (K) or calcium (Ca) was not successful, as the resulting cheeses exhibited a bitter taste.
More than 40 studies between the years of 1982-2012 dealing with the substitution of NaCl by potassium chloride (KCl) were analysed by Hoffmann in 2014. Flavour defects, in particular bitterness and loss of salty flavour, remained a problem. Further studies have reported on the reduction of sodium in a variety of cheeses including dry-salted Cheddar (e.g. Murtaza et al. 2014), Gouda (Ruyssen et al. 2013), Danish Samsoe (Sondergaard et al. 2015) and Tybo (Sihufe et al. 2018). In the production of Na-reduced Gouda (Ruyssen et al. 2013), one-third of NaCl in the brine was substituted by KCl. In addition to the usual starter cultures, Lactobacillus (Lb.) helveticus und Lb. paracasei were implemented as adjunct strains in an effort to improve taste. There were no significant differences in the chemometric results, but a trained taste panel determined significant differences in saltiness, bitterness, texture and preference between the reference NaCl-brined cheeses and the NaCl/KCl-brined cheeses.
In further studies, mixtures of NaCl/KCl and flavour enhancers (L-arginine, yeast and oregano extract) were used in typical Brazilian semihard probiotic Prato cheese (Silva et al. 2018a,b). Extensive analyses including quantitative descriptive analysis and temporal dominance of sensations (TDS) for sensory profiling were applied. Sodium reduction and the use of probiotic cultures may be an effective alternative for the production of a potentially functional cheese. Sodium reduction, addition of xylooligosaccharides, yeast extract and arginine were also explored by Ferrão et al. (2018) in requeijão cremoso processed cheese, another typical Brazilian cheese. These authors demonstrated that it was possible to manufacture a potentially prebiotic cheese with 50% Na reduction and 80% fat reduction that showed comparable physicochemical, rheological and sensory characteristics to the full-fat, regularsalt product.
Coalho, another popular Brazilian cheese with a mild aroma and a firm but soft texture, is traditionally manufactured with at least 2% NaCl to maintain its texture during heating. Costa et al. (2018) studied the effect of partial substitution of NaCl by KCl on the characteristics of Coalho and in a consumer test demonstrated that partial replacement of up to 50% NaCl by KCl may be a feasible alternative. Furthermore, São João cheese from Pico Island (Portuguese Azores) is a regionally highly consumed cheese with a high salt content and a moderate to intense aroma. Considering sensory, physicochemical and microbiological results, Soares et al. (2015) demonstrated that a 25% salt reduction in this cheese is feasible on an industrial scale as it was not detected by the regular consumer.
The ambitious aim of the present study was to reduce the Na content of film-ripened, semihard Edam cheese to ≤0.4 g Na/100 g (≤1 g NaCl/100 g), while retaining the typical quality and microbiological safety of this cheese. This was carried out by applying simple NaCl reduction and mineral salt substitution mixtures containing Na and K to reduce overall sodium content, as well as by implementing specific starter and adjunct cultures in an effort to improve taste.

MATERIALS AND METHODS
Materials Raw bovine milk for cheese production was obtained from the experimental farm (Schaedtbek, Germany) at the Max Rubner-Institut (MRI). The starter cultures implemented were obtained from Chr. Hansen (Nienburg, Germany). The species descriptions, as provided by the manufacturer, are outlined below. Further species clarification was not obtainable due to the trade secret of their recipes. The F-ES Easy-Set â FLORA TM C-1060 culture contained Lactococcus (Lac.) lactis subsp. cremoris, Lac. lactis subsp. lactis biovar diacetylactis, Lac. lactis subsp. lactis and an unspecified Leuconostoc species. F-DVS CR-550 was a mixed culture of Lactobacillus (Lb.) species and Lac. lactis subspecies, while F-DVS LH-32 contained only Lb. helveticus. Finally, F-DVS CR-BUTTERY01 contained Lb. paracasei, Lb. rhamnosus and Lac. lactis subsp. lactis.

Cheese production
Cheese was manufactured in a pilot plant at the MRI cheese laboratory. Time lapse, temperature settings, pH values and addition of ingredients are presented in Table 1. The raw milk was skimmed using a disc centrifuge (GEA Westfalia Separator Group, Oelde, Germany). A calculated amount of the separated cream was added to the skimmed milk to adjust the fat content (2.45-2.70 g/100 g) according to the protein content (3.30-3.55 g/100 g). The adjusted milk was pasteurised using an APV Rosista milk heater (300 L/h; Unna, Germany) and cooled. Further manufacturing was performed in a cheese vat. During warming of the milk, 0.02 g CaCl 2 /100 g and 0.015 g KNO 3 /100 g were added. The freeze-dried starter cultures were thawed in warm water (30°C) for 1 h prior to addition to the milk. mucorpepsin, produced by submerged fermentation with Rhizomucor miehei; Chr. Hansen, Nienburg, Germany) was used. After cutting and curd/whey treatment in the cheese vat, the curd was transferred into 16 cylindrical plastic cheese moulds with individual press covers (diameter and height 14 cm; MilkySky GmbH, Lauben, Germany). After three pressing cycles with increasing pressure (Table 1), the resulting loaves were individually weighed and marked. Subsequently, the loaves were distributed equally into two brine solutions of different composition per experiment. The addition of 0.13 g Ca/100 g brine (as CaCl 2 9 2 H 2 O) was recommended by a commercial dairy as the typical content in brine. One brine solution always contained 17 g NaCl/100 g brine and served as the classically salted Edam control cheese, whereas the second brine solution was adapted to the specific experiment, for example through the implementation of sub4salt â ( Table 2). The loaves lay between two perforated sheets of stainless steel so that they were continually covered with the brine. After varying brine retention times, the loaves were transferred to wooden boards and allowed to dry until the next morning. They were then packed into plastic film shrink bags for cheese (300 9 400 mm; IP Ingredients, Suederluegum, Germany), vacuum packed and sealed (PTHW-KV-410-T-1, Allpax, Lauben, Germany). Before ripening, the packed cheeses were immersed briefly in hot water to shrink the plastic film.
In addition to alterations in brine composition and retention times, an experiment in which the milk used for the production of sodium-reduced cheese was deliberately inoculated with 1 9 10 5 cfu/mL of Listeria (L.) innocua, to simulate contamination with the related pathogen L. monocytogenes during manufacturing, was performed in biological triplicates. Listeria innocua is recommended by the 'European Union Reference Laboratory for L. monocytogenes' for implementation as a surrogate for the pathogen L. monocytogenes in pilot plant studies based on comparable growth conditions (EURL 2019). All other technical parameters remained the same as indicated above. Finally, selected cheese variations were manufactured in the laboratory of a commercial dairy under conditions simulating industrial manufacturing. The starter cultures and microbial rennet were the same as before. Manufacturing conditions were adapted to the larger film-ripened loaves with a weight of about 15 kg. In particular, the retention period of the cheese loaves in the brine had to be extended considerably. Varying retention times in the brine resulted in low salt (short retention time) and regular salt (long retention time) variants for both a NaCl brine and a mineral salt brine using sub4salt â . The exact brine retention times of these 15 kg loaves were not revealed by the commercial dairy laboratory.

Chemical and physical analyses
The development of pH was monitored during cheese manufacture and ripening by using an insertion pH electrode (SenTix â Sp, WTW, Weilheim, Germany).
For analysis of the ripened cheese, 3.5 cm thick slices were cut from the middle of two cheeses of each batch. Approximately 1 cm of the outer cheese surface (rind) was removed and the remaining core was grated prior to analysis. After six weeks of ripening, the pH and the content of dry matter, fat, protein, as well as NaCl in cheese were determined according to German standard methods (VDLUFA 2003). Technical triplicates of the grated core of two loaves from the same brining batch were analysed.
Na and K were analysed using ion-sensitive electrodes according to Rabe (1983) in a Konelab 20i (Thermo Fisher Scientific, Waltham, MA, USA). Homogenisation of the cheese samples using Tris buffer (pH 7.8; Tris and HCl from Merck, Darmstadt, Germany) was carried out in a Stomacher blender (Stomacher Mix 1, Kleinfeld Labortechnik, Gehrden, Germany) for 6 min at maximum speed and subsequently tempered at 80°C in a shaking water bath.
Technical quadruplicates of cheese samples from two loaves of the same brining batch were analysed. Before analysing the cheese loaves of each experiment, calibration curves were constructed considering the measuring range of the electrodes. The calibration curve for Na (nine measured values) ranged from 20 to 200 mmol/L (%0.46 g Na/100 g), and the curve for K (seven measured values) ranged from 2 to 20 mmol/L (%0.078 g K/100 g). The calibration solutions used were NaCl or KCl, dissolved in Tris buffer. Additionally, recovery tests of each analysed cheese were performed. Before homogenisation of the cheese using Tris buffer in the Stomacher blender, adequate amounts of NaCl and KCl were added and their recovery was checked. These tests confirmed the applicability and consistency of the method. The content of Na in cheeses containing only NaCl was also determined according to German standard methods (VDLUFA 2003), and the Na results were confirmed using an ion-selective electrode.

Analysis of peptides
For liquid chromatography-mass spectrometry (LC-MS) analysis of peptides in cheese, 2.5 g of cheese was homogenised in 50 mL of 0.1 M sodium phosphate buffer (pH 6.7) in a Stomacher blender (Stomacher Mix 1, Kleinfeld Labortechnik, Gehrden, Germany) for 6 min at maximum speed and room temperature. The cheese extract was adjusted to pH 4.6 with 4 M HCl and centrifuged at 4000 g at 4°C for 30 min. A volume of 0.5 mL was removed from the supernatant and purified using a 200 mg solid phase extraction column (Strata-X Polymeric SPE, Phenomenex â , Aschaffenburg, Germany). Each cheese sample was extracted in duplicate, and each extract was analysed with LC-MS in triplicate and compared to a sample of commercial Edam cheese used as a control sample.
The determination of peptide profiles was performed with LC-MS using an HPLC system (UltiMate TM 3000 RSLCnano system, Thermo Fisher Scientific, Bremen, Germany) and an LTQ XL TM linear ion trap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). For chromatographic separation, a PepSwift TM monolithic capillary LC column (200 µm 9 25 cm nanoViper TM column; Thermo Fisher Scientific, Bremen, Germany) with a PepSwift TM monolithic guard column (200 µm 9 5 mm nanoViper TM column, Thermo Fisher Scientific, Bremen, Germany) was used at a flow rate of 1 µL/min and a column temperature of 40°C. The mobile phases consisted of bi-distilled water with 0.1 mL/100 mL formic acid, and 80 mL/100 mL acetonitrile with 0.1 mL/100 mL formic acid, respectively, and applied in a linear gradient between 2 and 50% acetonitrile for 40 min. Samples were kept at 5°C until injected, and the injection volume was 1 µL. The MS experiments were performed in the positive ion mode using nanospray ionisation with a spray voltage of 2 kV and a capillary temperature of 200°C. A data-dependent scan with fragmentation of the three most intense ions (activation type = CID [35 eV]) and an isolation width of 2 Da was performed. MS data interpretation was performed with a Proteome Discoverer 1.4 (Thermo Fisher Scientific, Bremen, Germany) using the search algorithms SEQUEST and MASCOT without enzyme specification. The UniProtKB database (The UniProt Consortium 2019) restricted to Bos taurus was used for identifying peptides according to their MS values (Li et al. 2019), and the peptides identified by the algorithm were manually verified. Briefly, the identified peptide sequences were verified by comparing the measured fragment spectra of the peptides to the theoretical peptide fragments. The spectra have to match at least five y-, b-or a-ions of the theoretical peptide fragments. Furthermore, all major fragment masses of the spectra with intensities greater than 10-20% of the maximum intensity in the MS/MS must match theoretical peptide fragments. The peak area of the selected precursor ion of each peptide was calculated relative to the peak area of the same peptide in a commercial Edam cheese sample.

Microbiological analyses
For microbiological testing, samples were taken at five time points during manufacturing in the MRI cheese laboratory: after inoculation of the milk with starter cultures, from the curd directly before pressing, and after one, three and six weeks of ripening. Biological triplicates of each cheese production were analysed, and each bacterial count was measured in technical triplicates.
For testing, 10 g of cheese were placed in a BagFilter â 400 P (Interscience for Microbiology, Saint Nom, France) laboratory blender bag with a <250 µm lateral filter, to which 90 mL of 2 g/100 mL sodium citrate solution (pre-warmed to approx. 30°C) was added. The bag was placed in a BagMixer â (Interscience for Microbiology, Saint Nom, France) laboratory blender for 2 min at maximum speed. One mL of the filtered supernatant was diluted in a tenfold dilution series. Volumes of 100 lL of appropriate dilutions were spread-plated onto M17 agar (Terzaghi and Sandine 1975) to determine the bacterial count (cfu/ g = colony-forming units per gram) of lactococci and onto MRS agar (VWR International GmbH, Darmstadt, Germany) to determine the total number of lactobacilli (MRS pH 5.7) or leuconostocs (MRS pH 5.4). For Listeria counts, the samples were spread-plated onto ALOA agar (Ottaviani et al. 1997).
In addition, all samples taken after six weeks of ripening were tested for possible contamination with either enterobacteria by using Violet Red Bile Dextrose medium (VRBD, Merck, Darmstadt, Germany), enterococci by using Kanamycin Esculin Azide Agar (KAA, Merck, Darmstadt, Germany), yeast or moulds using Yeast Glucose Chloramphenicol Agar (YGC, VWR International GmbH, Darmstadt, Germany) and pseudomonads using Cetrimide Agar (CFC medium as described by Merck (2010)  Water activity (a w ) was measured at all of the above-mentioned time points using a HygroLab C1 bench-top indicator with digital a w humidity-temperature probes (Rotronic Measurement Solutions, Bassersdorf, Switzerland).

Sensory analyses
A pool of 40 panellists (25 female/15 male) was trained at the MRI in Kiel for the perception of cheese odour and taste. At least 20 panel members participated in each sensory test. Descriptive sensory analysis (profile testing) according to DIN EN ISO 13299:2016-09 (2016 was performed. Odour was tested for intensity, sour and off-odour, while taste included intensity as well as bitter, salty, sour, metallic and off-taste. The perception of each of these attributes was scored between 0 (none) and 4 (very strong).
Additionally, a consumer test according to DIN EN ISO 11136:2017-10 (2017 with 64 participants (employees of the MRI in Kiel and Hamburg) was carried out on the cheeses produced in the cheese laboratory of the dairy under simulated industrial conditions. These cheeses were assessed by 35 women and 29 men, of which 26 were ≤40 years old while 38 were >40 years old. Here, the only criterion tested was 'preference' (measured using a ninepoint hedonic scale ranging from 'dislike extremely' to 'like extremely').

Study replications
Following preliminary experiments, three biological replicates for Na-reduced cheese production with either a NaCl or a mineral salt brine (sub4salt â ) were analysed. Three independent experiments were performed for cheese with and without L. innocua co-inoculation. The arithmetic mean was calculated for each experiment, and error calculations were indicated as standard deviation or standard error of the mean in microbiological analyses. Further details are indicated in the table and figure legends.

RESULTS AND DISCUSSION
Chemical and physical properties Initially, film-ripened Edam cheese was produced in the MRI cheese laboratory using a range of mineral salts in the brine with different retention periods of the pressed loaves ( Table 2). The goal of these preliminary tests was to obtain benchmarks of Na and K content in the resulting cheeses. All starter cultures with the exception of F-DVS BUT-TERY01 were implemented in these initial cheese production experiments.
The cylindrical loaves of Edam cheese had an average mass of 1 kg, with their dry matter ranging between 52.1 and 54.7 g/100 g. An average pH value of 5.40 AE 0.08 after cheese ripening was measured. The results of Na and K contents analysed by ion-sensitive electrodes are presented in Table 2. The content of Na was between 0.24 and 0.77 g/100 g in cheese. Nine different salt/time combinations in the brine resulted in cheeses with ≤0.4 g Na/100 g, corresponding to the aim of this study. Fourteen combinations had ≤0.3 g K/100 g and, as sensory tests showed that K content exceeding this value in ripened cheese presented a bitter taste (results not shown), brine/ retention time combinations resulting in >0.3 g K/100 g were not suitable for Na reduction in Edam cheese and were disregarded for further experiments. Cheeses that were salted in pure NaCl brine contained 0.05-0.08 g K/100 g. These data correspond with the average value of 0.067 g K/ 100 g in semihard cheese stated in the Federal Ministry of Food and Agriculture ( 2017).
Evaluation of the preliminary experiments allowed for the selection of seven variants of brine/retention combinations, six with mineral salt combinations and one simply with reduced NaCl (Table 3). The chosen combinations of NaCl and the commercial mineral salts in the brine are based on their Na and K contents. Our aim was to produce cheeses with low Na content and a K content of ≤0.3 g/100 g. Sub4salt â was applied in the brine without additional NaCl, as the proportion of Na and K in this product resulted in the desired content of these minerals in the resulting cheeses ( Table 2). Each of the mineral salt combinations was tested against a pure NaCl cheese, always immersing eight of the pressed loaves in the control brine (17 g NaCl/100 g brine, 45 min) and eight loaves in the brine with the different mineral salts. To increase aroma development, the starter cultures were supplemented with F-DVS BUTTERY01 (recommended by Chr. Hansen, Nienburg, Germany). The results outlining Na, K and dry matter in the cheese are presented in Table 3. The average dry matter for all combinations was between 53.68 and 55.83 g/100 g. Therefore, the minimum dry matter of 53% for German standard Edam cheese with 40% fat in dry matter was always exceeded. This desired effect was achieved by a more intensive curd treatment, which resulted in higher whey drainage during cheese manufacture compared to the preliminary experiments. An effect of the higher dry matter was that the Na content was continuously lower than 0.4 g/100 g. The average pH value of all cheeses analysed 24 h after the addition of starter cultures was 5.35 AE 0.05, which was 0.03 higher than in the preliminary tests. After ripening, the pH had increased to 5.51 AE 0.03, which was 0.12 higher than in the preliminary tests. This increase in pH value may be explained by proteolysis influenced by the additional adjunct starter culture F-DVS BUTTERY01.
The Na content in the cheeses that were immersed in the pure NaCl brine was 0.24-0.34% (Table 3), a result of the slightly higher dry matter, as mentioned above. For example, a dry matter of 55.6 g/100 g resulted in only 0.24-0.27 g Na/100 g and a K content of 0.06-0.07 g/100 g, confirming preliminary results. The cheeses salted in brine containing mineral salts had 0.20-0.27 g Na/100 g and 0.13-0.20 g K/100 g. The aim for <0.4 g Na/100 g was also achieved for the mineral salt cheeses.
Of the seven variants of brine/retention combinations, one of the mineral salt variants (sub4salt â ; brine retention time of 1.5 h), the simple NaCl reduction variant (short brine retention time of 1 h = low salt, extended by 15 min for a slightly higher Na content) and a reference using the NaCl brine without shortened retention time (long brine retention time of 4 h = regular salt) were chosen mainly for detailed microbiological analyses in a further set of cheese production experiments (Table 4). These experiments used identical processing conditions as before. The sub4salt â mineral salt mixture was chosen because its composition was suitable to achieve the Na content aims of this project. The preliminary tests suggested sub4salt â could be applied without the need for additional NaCl in the brine, contrary to all other substitution salts tested. However, clear sensory advantages between any of the mineral salt products tested Table 3 Effect of brine composition and retention time on the Na, K and dry matter contents of seven selected Edam variants (n = 1 brine each, 2 cheeses per brine).

Mineral salts in brine, retention time
Na, K and DM contents (g/100 g) in preliminary experiments were not noticed. The dry matter of these manufactured cheese loaves was between 55.43 and 56.06 g/100 g as a result of the protein content (≥3.51 g/ 100 g) and corresponding adjusted fat content (≥2.68 g/ 100 g) being at the upper limit of the milk used for cheese production. The NaCl cheeses with short and long retention times contained 0.29 and 0.43 g Na/100 g, respectively, and both contained 0.06 g K/100 g on average. The cheese with sub4salt â had 0.21 g Na/100 g and 0.16 g K/100 g on average.
Cheeses produced in the laboratory of the commercial dairy had pH values between 5.46 and 5.53.
As mentioned previously, the cheese loaves had a weight of about 15 kg and required a considerably prolonged brining time in contrast to the 1 kg loaves manufactured in the MRI laboratory. Nevertheless, the terms 'short' and 'long' retention times, resulting in 'low' and 'regular' salt content, were retained as the resulting cheeses were comparable to those manufactured in the MRI pilot plant (Table 4). The two variants immersed in the NaCl brine had an average dry matter of 55.79 g/100 g (short retention time = low salt) and 57.56 g/ 100 g (long retention time = regular salt), respectively, while the cheeses immersed in sub4salt â brine had average dry matter of 54.16 g/100 g (low salt) and 54.46 g/100 g (regular salt) ( Table 4). The additional variant with extended retention in the sub4salt â brine (regular salt) was produced to get additional information on resulting Na and K contents with minimum effort. The NaCl-brined cheeses of the low and regular salt variations had a Na content of 0.33 and 0.56 g/100 g and a K content of 0.08 and 0.07 g/100 g, respectively. The sub4salt â brined cheese variants with low and regular salt contained 0.29 and 0.44 g Na/100 g, and 0.19 and 0.27 g K/100 g, respectively. Both the mineral salt substitution method and the reduced NaCl method could be applied successfully to production in the larger 15 kg format used in the commercial dairy research laboratory. However, the Na and K values measured in the cheese experiments performed in the MRI pilot plant were minimally lower than those measured in the samples produced in the commercial cheese dairy laboratory. This discrepancy could be due to slightly variable diffusion rates based on the different loaf weight (1 kg at the MRI and 15 kg in the commercial dairy).

Effect of salt reduction on proteolysis
Analysis of the peptides in Na-reduced Edam by LC-MS revealed a complex peptide pattern depending on the maturation stage of cheese samples, but showed no effect of brine composition or brining time. Bitterness in cheese is commonly associated with the accumulation of hydrophobic peptides formed through the hydrolysis of a s1 -and b-caseins (Baptista et al. 2017). Two large, hydrophobic peptides released from b-casein (b-CN) could be identified by their fragmentation patterns of the MS 2 -spectra. The peptides b-CN f193-209 and b-CN f194-209, the latter of which has been shown to be extremely bitter (Visser et al. 1983;Ard€ o et al. 2017), could be quantified relative to a commercial Edam cheese sample (Figure 1). The highest amounts of both peptides were found after one week of cheese maturation, declining during further cheese ripening and this was independent of brine composition. No effect of salt reduction on the generation of these bitter peptides was found, but further analysis of the influence of salt reduction on proteolysis and the role of taste enhancing peptides (Harth et al. 2018) should be done in further studies.

Growth of starter cultures
Chemical and physical analyses were complemented by microbiological studies to determine the effect of different brine composition/retention times (simple NaCl reduction Table 4 Chemical and physical properties of the brine composition and retention times for Na-reduced Edam produced in the MRI pilot plant for detailed microbiological analyses (n = 3) and in the research laboratory of a commercial dairy (n = 1).
Mineral salts in brine, retention time Na, K and DM contents (g/100 g) and sub4salt â substitution) on the growth of starter cultures during manufacturing and ripening of Edam cheese in the MRI cheese laboratory. The bacterial count of Lactococcus species increased rapidly from approx. 5 9 10 7 cfu/mL at inoculation to 5 9 10 8 cfu/g in the curd, further increasing to 1 9 10 9 cfu/g after one week, before decreasing to approx. 5 9 10 8 cfu/g and 3 9 10 7 cfu/g at three and six weeks, respectively (Figure 2a). The number of Leuconostoc species increased slightly from 1 9 10 5 cfu/mL at inoculation to approx. 3 9 10 5 cfu/g in the curd, before rising to and remaining between 5 9 10 6 cfu/g and 1 9 10 7 cfu/g for the remainder of the ripening time ( Figure 2a). Finally, Lactobacillus species counts developed from 6 9 10 6 cfu/ mL at inoculation to approx. 2 9 10 7 cfu/g in the curd before reaching, and remaining at, 1 9 10 8 cfu/g from one to six weeks of ripening (Figure 2a). Both the sub4salt â and the low NaCl as well as the regular NaCl cheeses showed similar bacterial growth in cheese. The progression of microbial growth of the starter and adjunct cultures during fermentation and ripening in Na-reduced Edam followed a typical development for Edam as documented in previous studies (Wachowska 2011;Ruyssen et al. 2013;Porcellato and Skeie 2016). As Edam cheese is mainly manufactured from pasteurised milk, nonstarter lactic acid bacteria (NSLAB) originating from the raw milk are of little significance. The interaction of starter lactic acid bacteria and NSLAB was reviewed by Blaya et al. (2018), who emphasised that the main source of NSLAB in cheese is the raw milk. In addition, bacterial counts for Lactococcus, Lactobacillus and Leuconostoc species showed a similar progression both with and without co-inoculation with L. innocua (Figure 2b). The Listeria counts showed approx. 3 9 10 5 cfu/ mL at inoculation, rising to 8 9 10 5 cfu/g in the curd, reaching a maximum at approx. 1 9 10 6 cfu/g after one and three weeks, before decreasing slightly to 7 9 10 5 cfu/ g after six weeks. The presence of L. innocua did not affect starter LAB growth, and the reduced sodium conditions did not lead to any substantial change in bacterial growth patterns of L. innocua. As L. innocua was implemented as a surrogate for L. monocytogenes based on its comparable growth conditions, it can be inferred that the growth of the pathogen L. monocytogenes would not be affected by the reduced Na conditions tested in this study.
After six weeks of ripening, selective agar tests for enterobacteria, enterococci, yeasts, moulds and pseudomonads showed bacterial counts of <1 9 10 2 cfu/g (limit of detection), regardless of the brining conditions of the cheese samples, suggesting that the microbial quality of the Na-reduced Edam produced in this study was not compromised by these spoilage and potentially opportunistic pathogens. In addition, the a w values of sodium-reduced Edam (both mineral salt substitution and simple NaCl reduction) were similar to standard sodium cheeses after six weeks of ripening (Table 5). The a w values always exceeded 0.97, and the sodium content in the aqueous cheese phase was between 0.48 and 0.98 g/100 g water in cheese. As the preservative effect of NaCl lies in its ability to reduce a w (Guinee and Fox 2017), our findings suggest that the NaCl content of the sodium-reduced Edam produced both with and without co-inoculation with the L. monocytogenes surrogate L. innocua did not affect preservation of these cheeses. However, further testing for additional pathogens is necessary.

Sensory characteristics
Following preliminary experiments, the cheeses produced in seven variants with different brine/retention combinations (Table 3) were evaluated by the MRI sensory panel considering the defined criteria. Additionally, commercial Edam cheese with 53.95 g/100 g dry matter, 40% fat in dry matter, 0.82 g Na/100 g and a pH of 5.32 was also organoleptically tested. This Edam originated from the same commercial dairy where the cheeses simulating industrial manufacturing were produced. The loaf was pressed in a bread form, had a weight of about 3 kg and was ripened for six weeks. The Na content corresponding to more than 2 g NaCl/100 g approximately represents the upper limit of Na in commercial German Edam cheese. The commercial Edam used for comparison was saltier and less bitter, but also sourer than all sodium-reduced cheeses produced (data not shown). All other variants showed no differences in any criterion. This trend was also observed in the simple NaCl reduction cheeses, supporting the notion that salt masks bitter taste occurring in cheeses of standard sodium content. Visser et al. (1983) found that the b-CN f193-209 peptide resulted in a very bitter taste and a study by Møller et al. (2013) in Cheddar cheese reported that the release of this peptide was Na-sensitive. However, our results showed no difference in the release of this peptide in relation to reduced Na content ( Figure 1). Additional bitter peptides and other causes must therefore be investigated in further studies. Based on the preliminary results, the sub4salt â and the low NaCl cheeses were further organoleptically tested and showed similar trends to the preliminary experiments (data not shown).
To simulate industrial processing, cheese blocks of about 15 kg passed through a NaCl brine or through a sub4salt â brine for two different retention times, producing cheeses with low and regular salt content (Table 4). The Na and K contents and the sensory profile analysis after cheese ripening are presented in Figure 3. The intensity of salty taste reflected the analysed Na content of these cheeses, which ranged from 0.29 to 0.56 g/100 g. From these results, it is evident that the intensity of salty taste corresponds to the total taste intensity. Off-odour was only perceived in cheeses that passed through the sub4salt â brine and these cheeses also presented a more bitter taste. In contrast, a difference in the intensity of metallic taste was not detected in all four cheeses (Figure 3). The bitter taste possibly overlapped other forms of off-taste so that they were not perceived by the Figure 2 Lactic acid bacterial counts of Lactococcus, Leuconostoc and Lactobacillus species during (a) standard cheese production and ripening (n = 3), and (b) during cheese production co-inoculated with 1 9 10 5 cfu/mL Listeria innocua (n = 3). Samples were taken directly after inoculating the pasteurised milk with starter cultures, from the curd immediately prior to pressing and from cheese samples after one, three and six weeks of ripening at 13°C. Depiction: arithmetic mean (n = 3) AE standard error of the mean. The standard error of the mean for all samples was ≤0.004; low NaCl = 0.29 g Na/100 g; regular NaCl = 0.43 g Na/100 g; sub4salt â substitution = 0.21 g Na/100 g.
panel. As off-flavour is an interaction of off-odour and offtaste, it can be noted that the cheeses passing through the sub4salt â brine developed such a flavour. These findings agree with the bitterness and off-flavours originating from the addition of potassium that have been reported in a previous study focussing on partial substitution of NaCl with KCl in the related, Dutch-type cheese Gouda (Ruyssen et al. 2013). Many studies on salt reduction in various cheese types, as well as other foods (e.g. fish products), have shown that the replacement of NaCl with KCl is possible up to about 30% (Hoffmann 2014;Giese et al. 2019). The mineral salt substitution product chosen for further analyses in this study (sub4salt â ) claims to allow for a 35% sodium reduction compared to conventional salt (Scholten 2007).
The same cheese variants were also tested for preference by a consumer panel without sensory testing expertise (Figure 4). The results of the preference test were graded between 'like slightly' and 'like very much' for all Na-reduced and regular samples. Although there was quite some discrepancy among individual evaluators within the consumer panel, these findings give an indication of consumer preference of the Na-reduced Edam samples among the general population. We are aware that the tests with the MRI sensory panel and with the consumers were only a starting point for more detailed experiments and analyses. Da Silva et al. (2014) determined the potency and equivalence of seven different salt substitutes in cream cheese, using magnitude estimation and TDS. The reference cream cheese for salty taste contained 1.0% NaCl. This corresponded to 1.2% NaCl low salt, 0.33 g Na/100 g, 0.08 g K/100 g NaCl regular salt, 0.56 g Na/100 g, 0.07 g K/100 g sub4salt® low salt, 0.29 g Na/100 g, 0.19 g K/100 g sub4salt® regular salt, 0.44 g Na/100 g, 0.27 g K/100 g odour/taste perception 0 none 1 weak 2 distinct 3 strong 4 very strong Figure 3 Sensory profile of ripened Edam cheese produced in 15 kg loaves at the research laboratory of a commercial dairy, displaying different brine composition (NaCl and sub4salt â ) and retention times (low salt and regular salt) of the cheese loaves. Depiction: arithmetic mean (n = 22) AE standard error of the mean.
KCl, 2.5% MgCl 2 or 2.98% potassium phosphate. The TDS analysis revealed that the seven salt substitutes resulted in other tastes in addition to salty taste, including significant sour and bitter tastes. Therefore, additional TDS experiments with Na-reduced Edam cheese are necessary, especially since ripened cheese is a more complex matrix than cream cheese.

CONCLUSIONS
The present study showed that through the change of brine composition and brining times, combined with adjustment of the implemented starter cultures, production of a sodiumreduced Edam cheese with <0.4 g Na/100 g was possible.
Analyses during production and after six weeks of ripening showed that the desired Na and K values in sodium-reduced Edam cheese were achievable without compromising the microbial quality of the cheeses and that there was no effect of salt reduction on the generation of the analysed bitter peptides. However, despite a considerable reduction of Na being technologically possible, the mineral salt substitution cheeses displayed a bitter taste typical for potassium. The addition of adjunct cultures for additional aroma production through increased proteolysis and the generation of volatile compounds did not suffice to prevent the detection of off-flavours. Simple NaCl reduction also resulted in Edam cheese that was bitterer than a commercial cheese with 0.8 g Na/100 g, in comparison with which it was also less salty and less sour. Both the mineral salt substitution method (sub4salt â ) and the simple NaCl reduction method could be applied successfully to production in the larger 15 kg format used in the laboratory of a commercial dairy. In general, Na reduction to <0.4 g/100 g is possible in Edam cheese; however, further work needs to be done to reduce off-flavours in Na-reduced cheese. cheeses, Angelika Thoß and Cornelia Voß for the chemical and physical analysis of products (Department of Safety and Quality of Milk and Fish Products), and Jennifer Grundmann (Department of Microbiology and Biotechnology) for microbiological analyses. Further thanks to the DMK Deutsches Milchkontor GmbH (Edewecht, Germany) and Chr. Hansen GmbH (Nienburg, Germany) for their technical support.  Figure 4 Preference of Na-reduced Edam, displaying different brine composition (NaCl and sub4salt â ) and retention times (resulting in low salt and regular salt cheese) and tested by a consumer panel (n = 64). Depiction: arithmetic mean (n = 26 for ≤40 years old and n = 38 for >40 years old) AE standard deviation.