To feed a growing human population and reduce the environmental impact of food production, we need sustainable, non-food alternatives for animal feed. The current geopolitical situation as well as the evident climate changes have revealed how vulnerable the global food system is. Increased domestic food production, together with using non-food resources in animal feed without seizing arable land, can reduce global food shortages and increase food security. The Nordic countries have a low degree of self-sufficiency and rely on imported feed (De Visser et al., Reference De Visser, Schreuder and Stoddard2014). In Norway, high-yielding dairy cows are usually fed a diet combining grass silage and barley-based concentrate, with soybean meal often used as the protein supplement (Nysted et al., Reference Nysted, Uldal and Vakse2020). However, soybeans also represent valuable protein sources for human consumption. Yeast, on the other hand, can be produced domestically by processing low-value, wood biomass (Øverland and Skrede, Reference Øverland and Skrede2017).
When modifying feed composition for dairy cows, the quality of milk and dairy products must be ensured. The lipid fraction contributes to distinctive characteristics, such as appearance, texture and flavour, satiety and nutritional properties of dairy products (McSweeney et al., Reference McSweeney, Fox and O'Mahony2020). Milk fatty acid (FA) composition is readily changed by the diet, primarily through alterations in FA supply (Truong et al., Reference Truong, Lopez, Bhandari and Prakash2020). The introduction of new protein sources will introduce distinct FA-profiles. Around half of the milk FAs are synthesized de novo, while the remaining half originates from the feed. However, due to rumen biohydrogenation or desaturases in the mammary gland, the milk FAs will be much more saturated and contain more trans and conjugated linoleic acid (CLA) intermediates than the feed FAs (Truong et al., Reference Truong, Lopez, Bhandari and Prakash2020). The effect of dietary protein on milk fat is generally considered to be small (Sutton, Reference Sutton1989). Nevertheless, dietary protein may have subtle effects on milk FAs by providing precursors for the synthesis of various odd- and branched-chain FAs (OBCFA; Cabrita et al., Reference Cabrita, Fonseca, Dewhurst and Gomes2003). Rumen bacteria are the major source of OBCFAs in milk, and changes in the relative abundance of specific bacterial populations in the rumen might lead to variations in the milk OBCFA-profile (Vlaeminck et al., Reference Vlaeminck, Fievez, Cabrita, Fonseca and Dewhurst2006). New protein sources also encompass additional substances, including carbohydrates, which can potentially impact the rumen microbiota and milk FA composition (Vlaeminck et al., Reference Vlaeminck, Fievez, Cabrita, Fonseca and Dewhurst2006).
This paper is based on a study of the effects of replacing soybean meal in concentrates with C. jadinii yeast in the diets of Norwegian Red (NR) dairy cows. Furthermore, as barley can be produced in Norway, and presently constitutes a major part of the concentrate, replacing soybean meal with additional barley was also evaluated. Previous results of the same study (Olsen et al., Reference Olsen, Vhile, Porcellato, Kidane and Skeie2021; Kidane et al., Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022) showed no effects of the different concentrates on feed intake, milk yield or the content of fat, lactose and protein in milk, nor on the sensory quality of cheese made from the same milk. Hence, we hypothesized that substituting soybean meal with C. jadinii yeast or incorporating more barley into the concentrate for NR dairy cows did not negatively influence the composition of milk FAs. The main goal of this study was to evaluate the effects of these concentrates on the rumen microbiota, FA composition and the sensory characteristics of the milk.
Materials and methods
Experimental design
The experiment was performed at the livestock production research centre at the Norwegian University of Life Sciences (NMBU, Ås, Norway) as described by Kidane et al. (Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022) and approved by the national animal research authority of the Norwegian Food Safety Authority (FOTS ID: 18038). In short, 48 Norwegian Red dairy cows in their early to mid-lactation (103 ± 33.5 d in milk, DIM) were allocated into three feeding treatments (n = 16). All animals were fed the same basal diet of grass silage (65% of dry matter intake, DMI), but different concentrates (35% of DMI). The concentrates were all based on barley as the main ingredient, but 7% of the barley quantity in the barley-concentrate (BAR, negative control) was replaced by either soybean meal (SBM, conventional control) or yeast microbial protein (YEA). The experiment lasted for 10 weeks. After 2 weeks of adaptation (control period) where all cows received the control diet (grass silage and SBM), the animals were allocated to experimental diets for a period of 8 weeks. More details can be found in the online Supplementary File.
Sampling and analysis
Samples from concentrates and silage were taken every week and stored at −20°C until FA analysis. An automatic milking system (De Laval, Lund, Sweden) was used with access every 6th hour. Representative milk samples were drawn from each cow from every milking from Monday to Wednesday weeks 2, 4, 6, 7 and 10. For each week, the samples were collected, pooled and mixed for each individual cow, and stored in darkness at −20°C until further analysis. For more information see online Supplementary File.
FA analysis was performed on milk samples from weeks 2, 6 and 10. The chemicals and procedures used for extraction, derivatization and GC-MS analysis are found in the online Supplementary File. In brief, lipids from 1.00 g of the samples were extracted using a modified version of the Folch method, as described by Devle et al. (Reference Devle, Ulleberg, Naess-Andresen, Rukke, Vegarud and Ekeberg2014). The lipids were transesterified to FA methyl esters (FAMEs) for identification and quantification using GC-MS. The FAME analysis was carried out as reported by Molversmyr et al. (Reference Molversmyr, Devle, Naess-Andresen and Ekeberg2022).
The sensory quality of pasteurized (63°C, 30 min) milk samples was evaluated by three professional assessors and classified approved or not approved. Unapproved samples had a distinct flavour deviation from normal milk.
Statistical analysis
The effects of concentrate and week on milk FA composition were analysed using RStudio with values from the SBM-group used as references and values from week 2 considered covariates.
The linear model used was:
$$y_{ijk} = \; {\rm \mu } + \; {\rm \gamma }_i + \; {\rm \tau }_j + \; K_k + \; E_{ijk}$$where Y ijk is the response variable, μ the overall mean, γi the fixed effect of the feed, τj the fixed effect of the week, K k the random effect of the cow and E ijk the error-term. For details see online Supplementary File.
The effects of the concentrates on the sensory quality were tested using the logistic procedure of SAS Enterprise Guide 7.1 (SAS, Cary, USA). The model used included the fixed effects of concentrates, week, parity (primiparous or multiparous), covariate (week 2), and the interaction between concentrate and week. For details see online Supplementary File.
Microbiota analysis
DNA extraction and 16S rRNA gene sequencing
The DNA extraction, sequencing and data treatment of the rumen are described in the online Supplementary File. In short, microbial DNA were extracted using the QIAamp PowerFecal Pro DNA Kit (Qiagen, GmbH, Hilden, Germany) and library preparation and 16S rRNA gene sequencing were processed by Eurofins Genomics (Ebersberg, Germany) or at NMBU on an Illumina MiSeq machine (Illumina, San Diego, CA, USA). The V3–V4 variable region of the bacterial 16S rRNA gene was amplified as described by Skeie et al. (Reference Skeie, Håland, Thorsen, Narvhus and Porcellato2019). The analysis of the demultiplexed paired-end raw reads was carried out using the DADA2 R package (version 3.11) (Callahan et al., Reference Callahan, McMurdie, Rosen, Han, Johnson and Holmes2016). Ribosomal Database Project (RDP) Naive Bayesian Classifier algorithm (Wang et al., Reference Wang, Garrity, Tiedje and Cole2007) implemented in DADA2 R package was used for taxonomy assignment to the ASVs using default settings and Silva prokaryotic SSU taxonomic training data formatted for DADA2 with species (release-138.1) (Quast et al., Reference Quast, Pruesse, Yilmaz, Gerken, Schweer, Yarza, Peplies and Glöckner2013). The filtered ASV table were pre-processed using the multivariate statistical framework mixMC (Cao et al., Reference Cao, Costello, Lakis, Bartolo, Chua, Brazeilles and Rondeau2016) in the mixOmics R package (version 6.19.4) (Rohart et al., Reference Rohart, Gautier, Singh and Lê Cao2017). Principle Component Analysis (PCA) was performed to identify the variation in the microbiota composition over time based on the diet groups. The taxa were agglomerated at the genus level and the lmmsDE function in R package lmms (version 1.3.3) was used to fit linear mixed effect model splines to perform differential abundance analysis.
Results
Fatty acid composition and sensory quality
In the feed, 28 FAs in the concentrates and 30 FAs in grass silage, were identified and quantified. The results are presented in Supplementary Table S1, while an overview of the distribution of SFAs, MUFAs and PUFAs in the different feeds are presented in Fig. 1a. The most abundant FAs in the concentrates were C16:0, C18:0, C18:1n-9cis9 and C18:2n-6cis9,12. Grass silage contained a higher amount of PUFAs and lower amounts of SFAs and MUFAs compared to the concentrates. The content of SFAs was lower in the YEA compared to the SBM and BAR, while the content of MUFAs and PUFAs were higher (P < 0.05; Fig. 1a). YEA contained less of the most abundant FA, C16:0, with 47%, compared to the SBM and BAR with 52–53%. The content of C18:1n-9cis9 and C18:2n-6cis9,12, on the other hand, was higher in the YEA compared to the SBM and BAR, with amounts of 32% of C18:1n-9cis9 and 11% of C18:2n-6cis9,12, compared to 29 and 8%, respectively. In addition, two of the less abundant FAs, C17:1n-8cis9 and C16:1n-7cis9, stood out with a substantially higher amount in the YEA compared to the SBM and BAR. The amount of C17:1n-8cis9 was 0.13% in the YEA and 0.02% in the SBM and BAR, and the amount of C16:1n-7cis9 was 0.28% in the YEA, while it was 0.13% in the SBM and BAR.
The distribution of saturated fatty acids (SFAs, blue), monounsaturated FAs (MUFAs, orange) and polyunsaturated FAs (PUFAs, grey) in grass silage and three concentrates: SBM, barley-based with additional protein from soybean meal; BAR, completely barley-based with no additional protein source replacing soybean meal and yeast; YEA, barley-based with additional protein from yeast (C. jadinii: a), and the same distribution of FAs in milk samples from dairy cows fed grass silage augmented with the three different concentrate feeds (b).
In milk, a total of 67 different FAs were identified and quantified. An overview of the milk FA composition can be found in Supplementary Table S2. The FAs were classified and grouped according to chain length, branching, and presence and position of double bonds (Table 1). The groups of FAs in the milk were not affected by the type of concentrate used, but there were differences due to week into the experiment for the class of very long-chained FAs (VLCFAs, >C20), trans-FAs, n-6 PUFAs and the ratio of omega-6 to omega-3 PUFAs (n-6/n-3) (noted with a superscripted e in Table 1). The VLCFA was the least abundant group of FAs (<0.2 g/100 g). The amount increased from week 2 to 6 for all feeding groups but decreased from week 6 to week 10. The amounts of trans FAs, n-6 PUFAs and the n-6/n-3-ratio decreased during the experimental period. The distribution of SFA, MUFA and PUFA in the milk were consistent among treatments and with time (Fig. 1b). This was also the case for the long-chain FAs (LCFA, C13–C20), which were the most abundant group (93.8 g/100 g on average).
The proportions (g/100 g of fatty acids (FAs)) of different classes of FAs in the milk from cows fed grass silage augmented with three different concentrates. SBM, barley-based with additional protein from soybean meal; BAR, completely barley-based with no additional protein source; YEA, barley-based with additional protein from yeast (C. jadinii)
a All cows were fed the SBM concentrate for two weeks in the control period.
b Standard error of the mean for the experimental period.
c SCFA, short chain FA (C1–C6); MCFA, medium chain FA (C7–C12); LCFA, long chain FA (C13–C20) and VLCFA, very long chain FA (>C20): The length of the FAs was classified according to directions from Christie (Reference Christie2023). BCFA, Branched chain FA.
d Standard error of the mean for the control period.
e Significantly affected by experimental week (P < 0.05).
Although there were no effects by the concentrates on the groups of FAs in milk, there were effects on individual FAs. An overview of the significantly affected FAs (P < 0.05), either by type of concentrate (measured using the SBM-group as reference) or week into the experiment, is given in Supplementary Table S3. There were in total 20 FAs that were significantly affected by the present feeding experiment. Eight of these were affected by type of concentrate, while 14 were affected by week into the experiment (C20:3n-6cis8,11,14 and an unidentified isomer (UI) of C17:1 were affected by both experimental factors). The contents of the significantly affected FAs in milk were below 2.0 g/100 g and these FAs were long-chained. Two FAs, C17:1n-8cis9 and an unidentified isomer (UI) of C18:3, were significantly influenced by the YEA, as compared to the SBM. The amount of C17:1n-8cis9 increased during the experimental period for all feeding groups but increased more in the YEA-group (0.138 to 0.163 g/100 g for the YEA compared to 0.133 to 0.153 g/100 g for the SBM). The amount of C18:3 (UI) increased in the YEA-group from 0.032 g/100 g at the beginning of the experiment to 0.043 g/100 g in week 6, before it decreased to 0.029 g/100 g in week 10. For the SBM-group, the amount decreased steadily during the experimental period from 0.036 to 0.026 g/100 g. Six FAs were significantly influenced by the BAR, as compared to the SBM. These were iso-C15:0, C16:1n-9cis7, C17:1 (UI), C18:1n-7cis11, C18:1 (UI) and C20:3n-6cis8,11,14, all below 0.7 g/100 g. The amount of iso-C15:0, C18:1n-7cis11 and C20:3n-6cis8,11,14 was stable during the experimental period in the BAR-group (0.206–0.204 g/100 g, 0.682–0.680 g/100 g and 0.043–0.046 g/100 g, respectively), while the amount decreased for the SBM-group (0.222–0.196 g/100 g, 0.692–0.639 g/100 g and 0.048–0.038 g/100 g, respectively). For C16:1n-9cis7, the amount increased for the BAR-group (0.129–0.136 g/100 g), while for the SBM-group the amount was stable (0.127–0.125 g/100 g). For C18:1 (UI), the amount increased from 0.346 to 0.365 g/100 g in the BAR-group, while the amount decreased in the SBM-group (0.361–0.343 g/100 g). The amount of C17:1 (UI) increased for both the BAR- and SBM-group, but the increase was higher for the BAR-group (0.080–0.104 g/100 g compared to 0.089–0.097 g/100 g). The FAs significantly influenced by only the week into the experiment were Iso-C14:0, C15:0, C16:1n-7cis9, C16:0–371115-tetramethyl, C18:1n-7trans11, C18:1n-5trans13, C11:0-cyclohexyl, C18:2n-6cis9,12, C18:2n-7cis9trans11, C21:0, C20:4n-6cis5,8,11,14, and C20:4n-3cis8,1114,17.
The sensory quality was not affected by the concentrates nor week into the experiment (data not shown).
Rumen microbiota
The rumen microbiota of the cow was not significantly affected by the concentrates. However, compositional differences were found over time (Fig. 2a). As shown by the first principal component (PC1), the rumen microbiota composition in week 10 was different from that of weeks 2 and 6.
Principal component analysis (PCA) of the microbiota of the rumen (a) and the centred log-ratio transformed abundance of different bacterial groups in the rumen (b) at 2, 6 and 10 weeks into the experiment. The three concentrates are labelled: S, SBM, barley-based with additional protein from soybean meal; B, BAR, completely barley-based with no additional protein source replacing soybean meal and yeast; Y, YEA, barley-based with additional protein from the yeast C. jadinii.
Several bacterial groups in the rumen (Fig. 2b) were found to be differentially abundant between the concentrate-groups over time according to the linear mixed effect model splines analysis. Noticeably, at the end of the experiment (10 weeks), the abundance of Lactococcus was higher in the rumen of the cows in the YEA-group, and the abundance of an unclassified genus affiliated with Paludibacteraceae was higher in the rumen of the cows of the BAR-group.
Discussion
In this study, grass silage and concentrate comprised 65 and 35% on average of the DMI for the dairy cows during the feeding experiment, respectively (Kidane et al., Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022). In addition, grass silage had a higher fat content than the concentrates (17% higher than SBM) and is, therefore, a substantially higher contributor to the milk FA composition than the concentrates. However, since the main hypothesis of this study was that the protein supply of the concentrates would not have adverse effects on the milk FA composition, the contribution of the concentrates will be emphasized further.
The replacement of 7% (DM basis) of soybean meal with yeast and additional barley has only had a minimal effect on the FA-profile of the concentrates, as the obtained FA-profile only revealed small differences when comparing the YEA with SBM and BAR. The FA-profiles primarily reflected the fat supplement used with calcium soaps of mainly C16:0 (47%), C18:0 (6%), C18:1 (37%) and C18:2 (9%), constituting 3.04–3.38% of the ingredients in the concentrates (Kidane et al., Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022). In addition, since barley constituted 48.9–55.4% of the ingredients, the FAs in barley also affected the FA-profile of the concentrates. Since the fat constituents of the protein source will contribute to the FA-profile to some extent, this explains the small difference found when comparing the FA-profile of YEA with the SBM and BAR, along with the small variation in the amount of calcium soaps added to the concentrates.
The small difference found in the FA composition between the concentrates are not reflected in the FA composition of the milk. The milk FAs that were significantly affected by the present experiment were long-chained. Longer-chained FAs are to a large extent determined by the content and composition of the feed (Truong et al., Reference Truong, Lopez, Bhandari and Prakash2020), indicating that the changes in these FAs in the present study are a result of the dietary fat. However, FA changes in the feed and FA changes in the milk are not correlated events since rumen processing affects the transfer of FAs from the feed to the milk (Truong et al., Reference Truong, Lopez, Bhandari and Prakash2020). The substitution of soybean meal with yeast only affected two of the less abundant (<0.2 g/100 g) FAs in milk (C17:1n-8cis9 and C18:3 (UI)). This is not surprising, since the 7% difference in composition of the concentrates only results in a 2.5% difference in total DMI between the dietary treatments, along with an even smaller difference in the FA-profile of the concentrates. The crude fat content of C. jadinii represented only 0.39% of the ingredient composition of the YEA-concentrate compared to 3.04% from calcium soaps (Kidane et al., Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022). Therefore, the calcium soap FAs would contribute more to the milk FA composition than the yeast FAs. Furthermore, rumen processing will affect the transfer of unsaturated FAs by biohydrogenation, resulting in more saturated FAs, whereas calcium soaps of unsaturated FAs are partially protected from this process (Palmquist et al., Reference Palmquist, Beaulieu and Barbano1993). The amount of C17:1n-8cis9, which was 6.5 times higher in the YEA as compared to the SBM concentrate, increased in the milk from the YEA-group as compared to the SBM-group, indicating that the yeast FAs had a minimal effect on the milk FA composition. Although, it might also be a result of changes in the rumen microbiota since rumen bacteria is the major source of odd- and BCFAs (OBCFAs) in milk (Vlaeminck et al., Reference Vlaeminck, Fievez, Cabrita, Fonseca and Dewhurst2006). However, we did not find any variations in the presence of different rumen bacteria across the different feeding regimens. Since there were only two of the less abundant milk FAs that were affected by the replacement of soybean meal with yeast, the overall impact on milk quality would be insignificant, which is in accordance with the sensory results, and the results on gross composition obtained by Kidane et al. (Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022), showing no differences due to dietary treatment. The composition did, however, change as the experiment progressed for all feeding groups, and there was a significant interaction effect of sampling day by treatment for the milk fat content, with YEA having a higher fat content in samples taken in week 7. The protein content was also marginally (non-significantly) lower in the BAR group, compared to the other two dietary groups.
Replacing soybean meal with barley had a greater impact on milk FAs compared to the substitution with yeast. The BAR contained lower amounts of crude protein and higher amounts of starch, as compared to SBM and YEA (Kidane et al., Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022), and this might explain the changes observed in the milk FAs from the BAR-group. A study by Cabrita et al. (Reference Cabrita, Bessa, Alves, Dewhurst and Fonseca2007) showed that protein and starch concentration influenced the milk FA profile affecting several of the mid- and long-chained FAs, including C17:1n-9cis8 and C18:1n-6trans16. Two of the FAs affected by the BAR in this study were unidentified isomers of C17:1 and C18:1, which might be the same as the ones Cabrita et al. (Reference Cabrita, Bessa, Alves, Dewhurst and Fonseca2007) found. Decreasing dietary crude protein has also been associated with lower amounts of OBCFAs (Vlaeminck et al., Reference Vlaeminck, Fievez, Cabrita, Fonseca and Dewhurst2006), however, in the present study the amount of iso-C15:0 was stable in the BAR-group, while it decreased in the SBM-group, and C17:1 increased more in the BAR- compared to the SBM-group. If we consider the FAs significantly affected by week into the experiment, there are a total of seven OBCFAs affected by the present experiment, which again might be explained by changes in the activity of rumen bacteria. We did find differences in the abundance of certain bacterial populations with week into the experiment. In addition, the change in amount of trans FAs might be a result of changes in activity of rumen bacteria, as a range of trans FAs are formed during rumen biohydrogenation and if these are not hydrogenated, they can be found in milk (Truong et al., Reference Truong, Lopez, Bhandari and Prakash2020). As changes in the rumen microbiota are gradual, it is plausible that more prominent changes would have occurred if the experiment had been conducted for a longer period. However, a common practice in feeding experiments is Latin-square crossover experiments, with a shorter period on each experimental diet compared to the 8-week duration used in this experiment. It is also important to note that the changes in milk FA composition with week into the experiment might be a result of the stage of lactation (Palmquist et al., Reference Palmquist, Beaulieu and Barbano1993), which would also explain the effect of sampling day on the gross composition obtained by Kidane et al. (Reference Kidane, Vhile, Ferneborg, Skeie, Olsen, Mydland, Øverland and Prestløkken2022).
Our results contrast partly with Sabbia et al. (Reference Sabbia, Kalscheur, Garcia, Gehman and Tricarico2012), who observed increased mid- and long-chain milk FAs when replacing soybean meal with yeast-derived microbial protein. Higher levels of yeast-derived protein were used in their study (3.41% of total DMI) compared to ours (2.45% of total DMI), suggesting a potential effect with greater inclusion. However, our study followed standard feeding regimes for NR dairy cows. Li et al. (Reference Li, Gao, Lv, Lambo, Wang, Wang and Zhang2023) and dos Santos et al. (Reference dos Santos, Signoretti, de Oliveira, da Silva, de Oliveira Alves Rufino, de Souza, Pinheiro and Neto2022) found no effect on milk FA composition when replacing soybean meal with high-oil pumpkin seed cake and peanut meal at higher inclusion levels (17–18% of DM), aligning with our findings. However, their studies had shorter treatment periods using Latin square designs, and there was a change in the relative abundance of ruminal bacteria in the study by Li et al. (Reference Li, Gao, Lv, Lambo, Wang, Wang and Zhang2023). On the other hand, Zang et al. (Reference Zang, Santana, Moura, Galvão and Brito2021) and Dokou et al. (Reference Dokou, Athanasoulas, Vasilopoulos, Basdagianni, Dovolou, Nanas, Grigoriadou, Amiridis and Giannenas2023) demonstrated substantial effects on milk FA profiles with alternative protein sources. Zang et al. (Reference Zang, Santana, Moura, Galvão and Brito2021) replaced soybean meal with okara meal (15% inclusion) in a Latin-squared design with 21-d periods, while Dokou et al. (Reference Dokou, Athanasoulas, Vasilopoulos, Basdagianni, Dovolou, Nanas, Grigoriadou, Amiridis and Giannenas2023) used 6.2% flaxseed and lupin inclusion in a 60-d trial. Thus, the type of protein source, inclusion level, and experimental duration are important considerations when studying milk FA profiles.
In conclusion, C. jadinii can substitute soybean meal as the protein source in concentrates with a 7% inclusion to NR dairy cows without any effects on rumen microbiota, and without compromising the FA composition or sensory quality of the milk. Incorporating more barley in the concentrate instead of soybean meal, only had a minor effect on the milk FA composition.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029924000281
Acknowledgements
The authors thank the Animal Production and Experimental Unit for cooperation with milk logistics, TINE SA for sensory analysis, and KBM for assistance with sampling and analyses. The study was done as part of the Foods of Norway project, funded by the Research Council of Norway, grant no. 237841/030. NMBU has funded parts of the PhD-project (grant no. 1205051073). Figures have been edited in BioRender.com. No conflicts of interest were declared by the authors.
Open access
