Metabolomics approach to evaluate a Baltic Sea-sourced diet for

May 30, 2017 - The diet-induced changes on metabolic profile in fish plasma, liver and muscle were studied relative to a fishmeal-based standard diet ...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/JAFC

Metabolomics Approach To Evaluate a Baltic Sea Sourced Diet for Cultured Arctic Char (Salvelinus alpinus L.) Ken Cheng,*,† Elisabeth Müllner,† Ali A. Moazzami,† Hanna Carlberg,‡ Eva Bran̈ nas̈ ,‡ and Jana Pickova† †

Department of Molecular Sciences, Swedish University of Agricultural Sciences, P.O. Box 7015, 75007 Uppsala, Sweden Department of Wildlife, Fish and Environmental Studies, Swedish University of Agricultural Sciences, 90183 Umeå, Sweden



S Supporting Information *

ABSTRACT: Aqua feeds traditionally rely on fishmeal as a protein source, which is costly and unsustainable. A new feed was formulated in the study with Baltic Sea sourced decontaminated fishmeal, Mytilus edulis and Saccharomyces cerevisiae, and given to Arctic char (Salvelinus alpinus) for ten months. The diet-induced changes on metabolic profile in fish plasma, liver, and muscle were studied relative to a fishmeal-based standard diet by using a 1H NMR-based metabolomics approach. Fish fed the test diet had higher content of betaine and lower levels of trimethylamine-N-oxide and aromatic amino acids in plasma or tissues, which were mainly caused by the diet. The metabolomics results are useful to understand the mechanism of lower body mass, smaller Fulton’s condition factor, and a tendency of less lipid content observed in fish fed the test diet. Thus, modifications on the dietary levels of these compounds in the feed are needed to achieve better growth performance. KEYWORDS: aromatic amino acids, Baltic Sea, betaine, metabolomics, myo-inositol, Mytilus edulis, Saccharomyces cerevisiae, trimethylamine-N-oxide



size.10 Baker’s yeast (Saccharomyces cerevisiae) has a high reproductive rate when culturing on nonfood substrates, but the high content of nucleic acids limits its dietary usage for terrestrial animals.11,12 A new fish feed was formulated by mixing Baltic Sea-sourced fishmeal (sprat Sprattus sprattus and herring Clupea harengus), blue mussel meal, and intact baker’s yeast as dietary protein sources for Arctic char. In our pilot study, Arctic char showed a taste preference to the new fish feed compared to a fishmealbased standard feed.13 However, Arctic char had less body weight gain after feeding on the new diet in a ten-month feeding trial.14 1 H NMR-based metabolomics is a reliable technology for quantitative and high-throughput analysis to get insights into the metabolic changes in organism.15−18 This approach has emerged as a field of interest in food and feed analysis.8,19,20 Thus, the objective of the present study is to investigate the diet-induced changes in metabolic profile in fish plasma and tissues (liver and muscle) by using 1 H NMR-based metabolomics to understand the underling mechanisms of the differences in weight gain between the dietary groups.

INTRODUCTION With the decline of capture fisheries production during the last decades, aquaculture has been responsible for the impressive growth in supply of fish for human consumption.1 It is reported that aquaculture production mounted to about 74 million tonnes in 2014, which provided almost half of total aquatic food for human consumption.1 Fishmeal is traditionally used as a protein source for aqua-feeds to fulfill the fish nutritional requirements, especially for carnivorous fish species. However, because of its unsustainability and increasing cost, the levels of fishmeal inclusion in aqua-feeds are decreasing.2 Thus, development and usage of alternative protein-based feeds for aquaculture are needed to meet the growing demand for fish products.3−5 Ingredients for replacement that do not compete with human food and do not threaten ecological sustainable would be good candidates. The Baltic Sea is one of the most polluted marine areas in the world with environmental problems like persistent organic pollutants contamination, and eutrophication caused by an excess of nitrogen and phosphorus.6,7 Thus, the nutrients from the Baltic Sea that are less valuable for direct human consumption can be reintroduced back into the food chain in the form of fish feed. We have shown that detoxified Baltic Seasourced fish materials reduced adverse effects on energy metabolism and hepatotoxicity in Arctic char (Salvelinus alpinus L.) and proved to be valuable feed ingredients.8 Deshelled blue mussel (Mytilus edulis) and intact baker’s yeast (Saccharomyces cerevisiae) having similar amino acid profile to fishmeal were found to be promising dietary protein sources for Arctic char in terms of growth and nutrient digestibility, when using them separately to replace 40% of fishmeal.9 Additionally, blue mussels having excellent nutrient uptake capacity is suggested to be a potential cleaner of the eutrophic Baltic Sea, but they are less interesting for human consumption due to their small © 2017 American Chemical Society



MATERIALS AND METHODS

Fish Feed Formulation. A fishmeal-based standard diet (control diet) and a novel diet originated from the Baltic Sea (test diet) were used in the present study (Table 1). The two diets were iso-nitrogenic and iso-energetic and manufactured at Laukaa Aquaculture station in Finland by the Finish Game and Fisheries Research Institute. The fishmeal in the control diet was originated from the Atlantic Ocean Received: Revised: Accepted: Published: 5083

March 3, 2017 May 27, 2017 May 30, 2017 May 30, 2017 DOI: 10.1021/acs.jafc.7b00994 J. Agric. Food Chem. 2017, 65, 5083−5090

Article

Journal of Agricultural and Food Chemistry Table 1. Feed Formulation (g kg−1 Feed); Proximate Composition of Crude Protein, Lipids, and Ash (% of Dry Feed); and Total Energy (MJ kg−1 Dry Feed) in the Fishmeal-Based Control Diet and Test Diet Derived from the Baltic Sea control a

Swedish Board of Agriculture. It has been approved by the Umeå Ethical Committee for Animal Experiments no. A62−10. Sampling. At the end of the experiment, final body weight and length of all fish remaining were measured for calculation of Fulton’s condition factor (K-factor;

332

259 267 73 49 46.7 18.7 6.8 22.9

× 100 ), which is a

measure of fish body mass index, to assess an individual fish’s growth and health. Fish were randomly and equally sampled from the tanks using tricaine methanesulfonate as anesthetic. The blood samples were collected from the caudal vasculature using 5 mL heparinized syringes and centrifuged (2200g for 5 min) immediately. Afterward, fish were sacrificed by a blow to the head. Tissue samples (liver and white muscle) were dissected and frozen in liquid nitrogen immediately. Samples were stored at −80 °C before analysis. 1 H NMR-Based Metabolomics Analysis. NMR Sample Preparation of Liver, Muscle, and Plasma. Frozen fish liver and white muscle samples (100 mg, n = 18 fish/diet, equally coming from 3 tanks/diet) were homogenized (Ultraturax T25, IKA, Staufen, Germany) in ice-cold methanol/chloroform (2:1, v/v, 3 mL) for 1 min and then sonicated in an ice-cold water bath for 30 min. After addition of 1 mL of ice-cold water and 1 mL of ice-cold chloroform, samples were centrifuged (1800g, 4 °C) for 35 min to achieve phase separation. The collected aqueous supernatant (the polar phase) of samples was dried using an evacuated centrifuge (Savant, SVC 100H, Techtum Instrument AB, Umeå, Sweden) and then redissolved in 520 μL of sodium phosphate buffer (0.135 mol/L, pH 7.0). The method applied was according to previous studies.8,22 Nanosep centrifugal filters (3 kDa, Pall Life Science, Port Washington, USA) were used to get rid of the residual proteins in samples.23 The filters were first washed six times with 500 μL of 36 °C Millipore water in a centrifuge (2000g, 8 min at 36 °C) to remove glycerol in the filters. Afterward, the aqueous liver and muscle samples were filtered via centrifugation (12 000g, 4 °C). Phosphate buffer (130 μL, 0.135 mol/L, pH 7.0), D 2 O (50 μL), and sodium-3(trimethylsilyl)-2,2,3,3-tetradeuteriopropionate (TSP-d4, 30 μL, 0.3 mmol/L, Cambridge Isotope Laboratories, Andover, USA) as internal standard were added to 390 μL of filtrate and analyzed with 1H NMR in 5 mm NMR tubes (Bruker Spectrospin Ltd., BioSpin, Karlsruhe, Germany). Plasma samples (60 μL, n = 24 fish/diet, equally coming from 2 tanks/diet since the samples from one tank each diet were destroyed) were defrosted first, placed in washed filter tubes, and then centrifuged at 13 000g at 4 °C for 30 min. Plasma filtrates (40 μL) were mixed with Millipore water (55 μL), sodium phosphate buffer (50 μL, 0.4 mol/L, pH 7.0), TSP-d4 (10 μL, 5.8 mmol/L, Cambridge Isotope Laboratories, Andover, USA), and D2O (15 μL) and then transferred to 3 mm NMR tubes (Bruker Spectrospin Ltd., BioSpin, Karlsruhe, Germany). NMR Sample Preparation of Feed. Feed samples (50 mg, n = 3) were soaked in 50 μL of water for 30 min before extraction. The methods of extraction and filtration were the same as for tissue sample preparation. The samples were measured in 5 mm NMR tubes (Bruker Spectrospin Ltd., BioSpin, Karlsruhe, Germany). NMR Spectroscopy Acquisition and Data Processing. The samples were analyzed by a 600 MHz Bruker NMR spectrometer using zgesgp pulse sequence (Bruker Spectrospin Ltd., BioSpin, Karlsruhe, Germany) at 25 °C with 128 scans for aqueous tissue and feed samples extracts and with 512 scans for plasma samples. 1H NMR spectra were recorded with 65 536 data points over a spectral width of 17 942.58 Hz. The acquisition time was 1.8 s and the relaxation delay 4.0 s. All NMR spectra were processed using Bruker TopSpin 3.1 software. The data were Fourier-transformed after multiplication by a line broadening of 0.3 Hz and referenced to internal standard peak TSP-d4 at 0.0 ppm. Each spectral baseline and phase was corrected manually. A total of 48 metabolites in liver, 42 metabolites in muscle, and 57 metabolites in plasma were identified according to the NMR Suite 6.1 library (ChenomX Inc., Edmonton, AB, Canada), the Human Metabolome Database, and previous literature.23−27 The concen-

test

−1

Feed Formulation (g kg Feed) Atlantic Sea fishmeal Baltic Sea fishmeal mussel meal yeast wheat meal wheat gluten soy protein concentrate fish oil rapeseed oil Composition (% of Dry Feed) crude protein crude lipids ash Total Energy (MJ kg−1 Dry Feed)

Body weight (g ) (Total body length)3 (cm)

216 212 253 131 50 71 47 47.4 19.4 7.8 23.0

Both diets additionally contained 15 g kg−1 mineral/vitamin premix, 5 g kg−1 titanium oxide, and 40 mg kg−1 astaxanthin.

a

(Raisioagro Ltd., Raisio, Finland) and the fishmeal in the test diet from the Baltic Sea (TripleNine, Esbjerg, Denmark). The deshelled blue mussel meal in the test diet was from the southwest of the Baltic Sea (Royal Frysk Muscheln GnbH Emmelsbukk-Hornsbull, Germany). The Baker’s yeast was cultured on molasses, ammonia, phosphorus, magnesium, and vitamins and then dried on a fluidized bed (Jästbolaget, Stockholm, Sweden). To suit the fish size throughout the experiment, the feed pellets were extruded in 3 mm and 4 mm with a twin-screw extruder (BC-45 model, Clextral, Creusot Loire, France). Lipid fraction for both diets consisted of regular commercial fish oil and regionally produced rapeseed oil (Raisioagro Ltd., Raisio, Finland), which was afterward added to the pellets using a vacuum coater (Pegasus PG-10VC, Dinnissen, Sevenum, Netherlands). The contents of crude protein, crude lipid, and ash (% of dry feed) and energy (MJ kg−1 dry feed) in the feeds (Table 1) were determined, according to the methods described previously.13 The composition of amino acid in feeds was reported by Carlberg.14 Experimental Design. The detailed experimental design is described by Carlberg.14 In brief, Arctic char from the selected strain Arctic superior21 were individually tagged with PIT-tags (passive integrated transponders, Biomark HPT12). Juvenile fish with body mass 32.7 ± 10.1 g (SD, n = 2970) were randomly divided into six tanks (3 tanks/diet and 495 fish/tank) and reared in a flow-through system (65 L.min−1) in 2 m diameter tanks with 5 m3 water at Aquaculture Centre North (Kälarne, Sweden). Fish were fed a commercial diet (Skretting Nutra MP 1.9 mm and Skretting Optime 1P 2.5 mm, Stavanger, Norway) to acclimatize themselves to the environment for four months before the feeding trial. The whole feeding trial lasted ten months. It started by giving the 1:1 mixture of experimental and commercial feeds (Skretting Optime 1P 2.5 mm, Stavanger, Norway) for one month to reduce the risk of a sudden appetite loss. Afterward, fish were fed the experimental diets for the following 9 months. Fish grew from 50.1 ± 0.25 g to 628.0 ± 3.99 g (SE) during the ten months. Feeding was made ad libitum by belt-feeders from 7 a.m. to 4 p.m. in a light−dark cycle mirroring the natural conditions. The water temperature was between 1.2 and 13.8 °C (±0.1 °C, February to October). Thinning of fish was conducted twice to ensure suitable biomass in tank and fish welfare, while a sufficient number of individuals were kept for statistical evaluation.14 The experimental performance was in agreement with laws and regulations for experiments with live animals, in accordance with the 5084

DOI: 10.1021/acs.jafc.7b00994 J. Agric. Food Chem. 2017, 65, 5083−5090

Article

Journal of Agricultural and Food Chemistry

Table 2. Content (μmol/g−1 Feed) of Betaine, Myo-inositol, and Trimethylamine-N-oxide (TMAO); Lipid Content (%); and Fatty Acid Composition (% Total Identified) in Fishmeal-Based Control Diet and the Test Diet Derived from the Baltic Sea (Mean ± SE; n = 3)

trations of metabolites were recalculated from the spectra after accounting for overlapping signals using NMR Suite 6.1 profiler against TSP-d4 as internal standard, and expressed in μmol/g for tissue and feed, and μmol/L for plasma. Lipid Analysis. Total lipids in white muscle samples (n = 3 threepooled fish/tank, 1 g) and feeds (n = 3, 0.3 g) were extracted in 15 mL of hexane/isopropanol (3:2, v/v) and 6.5 mL of Na2SO4 (6.7%), according to the method described previously.28 The lipid contents were determined by weighing the organic extracts after evaporation of the solvent. The fatty acid (FA) compositions were analyzed as fatty acid methyl esters (FAME), which were prepared using boron trifluoride and dry methanol including 15-methylheptadecanoate (C17:1, Larodan Fine Chemicals AB, Malmö, Sweden) as an internal standard. They were measured by gas chromatograph (CP-3800, Varian AB, Stockholm, Sweden) equipped with flame ionization detector (FID), a split-mode injector, and a fused silica capillary column (BPX 70, SGE; length 50 m, id. 0.22 mm, film thickness 0.25 μm).28 The column temperature was started initially at 158 °C for 5 min and then ramped to 220 °C for 30 min, which was held for 8 min. Helium was used as carrier gas and nitrogen as makeup gas. The temperatures of injector and detector were 230 and 250 °C, respectively. The FA peaks were identified by comparing the retention times with an external standard (GLC-68A, Nu-check-Prep, Inc., Elysian, Minnesota, USA) using Galaxie chromatography software version 1.9 (Varian AB, Stockholm, Sweden). Data Analysis. The univariate data analysis was conducted by using Statistical Analysis System 9.3 (SAS institute, Cary, NC, USA). Lipid and FA data in percentage were arcsine-transformed before analysis. Distributions of normality and homoscedasticity were checked using Anderson−Darling test (P > 0.05) and Bartlett’s or Levene’s test (P > 0.05), respectively. If they failed the tests, the original data were log-transformed and retested. The effects of the test diets on K-factor and metabolomics were evaluated using the model PROC MIXED, with diet as fixed factor and tank as random factor. The effects on lipid profile were evaluated using the model PROC GLM, with diet as fixed factor. If the data did not satisfy the test of normality or homoscedasticity before and after log-transformation, they were analyzed using Mann−Whitney test. A P-value < 0.05 was regarded as statistically significant for K-factor and lipid data analysis. To account for multiple testing in metabolomics data analysis, Bonferroni correction (αβ = 0.0009 for 57 metabolites in plasma, αβ = 0.0010 for 48 metabolites in liver, and αβ = 0.0012 for 42 metabolites in muscle, Supplementary Table 1 and Supplementary Figure 1) was applied. For multivariate statistics on the metabolic data, SIMCA-P software 13.0 (Umetrics, Umeå, Sweden) was further applied. With all variables Pareto-scaled, principal component analysis (PCA) models were used to get a first overview of data and to detect outliers by using Hotelling’s T2 (95% confidence internal, CI) and DModX (95% CI). A regression model, orthogonal partial least squares-discriminant analysis (OPLS-DA), was further applied to seek the discriminative metabolites by using OPLS-DA loading plots and variable importance for the projection (VIP) plots. The test of cross-validation ANOVA (CV-ANOVA) and overall cross-validation R2 were used to check validation of the OPLS-DA models. The variables with VIP > 1 and the differences between VIP and its corresponding jack-knife-based CI larger than zero (VIP-CI > 0) were considered discriminative.

control betaine 5.43 myo-inositol 6.10 TMAO 10.0 lipid 17.3 Fatty Acid Composition 14:0 5.01 16:0 13.6 18:0 2.18 ∑SAFAb 21.4 16:1n-7 5.06 18:1n-9 27.4 18:1n-7 2.69 20:1n-9 3.46 ∑MUFAb 39.4 18:2n-6 12.1 ∑n-6b 12.5 18:3n-3 5.03 18:4n-3 2.18 20:4n-3 3.52 20:5n-3 7.96 22:5n-3 0.90 22:6n-3 7.10 ∑n-3b 26.7 ∑PUFAb 39.2

Pa

test

± ± ± ±

0.04 0.22 0.41 0.48

22.9 2.10 0.01 17.7

± ± ± ±

2.08 0.22 0.01 0.37

0.0001 0.0002