Both trials used diets with radiolabeled [4-¹⁴C] cholesterol and phytosterols to assess the direct effect of increasing levels of phytosterols on uptake of cholesterol in the context of a single meal. A single meal of phytosterols was administered to be able to separate immediate effects of phytosterols in the intestinal lumen from possible long-term systemic effects when phytosterols are accumulated in the tissues. Adult male zebrafish were fed individually, moved to clean tanks, and then kept overnight to allow time for uptake (see Supplementary Figure 1 for a schematic view of the setup). Both positive and a negative controls were included in the trials. A negative control group, named CON, was included in each trial and fed a diet without added phytosterols. A positive control group, named EZE, received the control feed without added phytosterols and a pharmaceutical blocker of cholesterol uptake, ezetimibe, dissolved in the water (10 μM, Santa Cruz Biotechnologies). This concentration has been shown to reduce cholesterol uptake by ∼30% (Clifton et al., 2010) and to significantly reduce the whole body cholesterol in zebrafish larvae (Baek et al., 2012). The positive control (EZE) was thus the control to verify that changes in cholesterol absorption were measurable in the current experimental setup. Twenty-four hours after feeding, fish were euthanized to determine uptake and excretion of the labeled cholesterol using a scintillation counter. The two trials differed in some ways: in trial 1, a phytosterol blend was added to the diets in graded concentrations and fish were dissected for organ samples, while in trial 2 a pure fucosterol extract diet was added, only one concentration of sterols was used, and whole fish were used for the scintillation counts. Pure fucosterol was added to the feeds in trial 2 due to its already documented effects on lipid metabolism (Hoang et al., 2012) and due to it being of interest for aquaculture nutrition with the introduction of algal sources of nutrients such as algae oil and -meal to the diets. Trial 1 was therefore able to provide us with information on tissue distribution of absorbed cholesterol. The simplified experimental procedure in trial 2 without dissections led to fewer samples per fish and dramatically shortened the handling time for sampling each fish, allowing us to have a larger amount of fish per treatment group and thus enable detection of smaller effects of the dietary treatments on cholesterol uptake than in trial 1.

The basal diet was formulated using standard feed ingredients selected for their low sterol concentrations (Table 1). Ingredients were thoroughly mixed with MilliQ water and dried in an oven at 42°C for 120 h before milling into pellets of the desired size (<1.12 mm and >0.56 mm). Both [4-¹⁴C]-labeled cholesterol (Perkin Elmer, MA, United States) and non-labeled cholesterol (>99% pure, Sigma) were coated onto the pellets with an aimed radioactivity of 0.3 MBq (18^∗10⁶ dpm) per gram of feed, in accordance with Hamre et al. (2011). Experimental diets were made by coating aliquots of the basal diet with the desired amounts of sterols. Sterols were dissolved in ethyl ether, vortexed well with the feed, and left to dry under nitrogen for ∼24 h to remove all traces of the solvent. Phytosterol concentrations were determined on a molar basis relative to the amount of cholesterol (Table 2). In trial 1, three concentrations of a phytosterol blend derived from soybean and rapeseed oil (CardioAid^TM, Archer Daniels Midland Company, average molar weight: 410.3) were used to create increasing ratios of phytosterol:cholesterol (1:1, 2:1, and 3:1) (Table 2). In trial 2, a fucosterol extract (>98% pure, Santa Cruz Biotechnologies, molar weight: 412.7) diet was used in addition to the phytosterol extract in trial 1, both included in a 2:1 ratio with cholesterol (Table 2). The molar weight of the phytosterol extract was based on phytosterol analyses of the product (Supplementary Table 1), performed as described in 2.3 Feed analyses.

Samples from the feeds were taken to measure final radioactivity (3 × 10 mg) and phytosterol concentrations (one sample of ∼200 mg per feed). Total fat was estimated by extracting lipids using heptane combined with HCl (European Economic Community [EEC], 2009). Crude protein was assessed by measuring total nitrogen and using 6.25 as multiplication factor (The Association of Official Analytical Chemists [AOAC], 1995). Sterols in the feeds were analyzed with a GC-FID setup and quantified using 5-β-cholestan-3-α-ol as an internal standard as described by Liland et al. (2013) adapted from Laakso (2005).

Since lipid metabolism may be very different between males and females (Meguro et al., 2015), only male wild type zebrafish were used in both trials to reduce variation. In trial 1, 48 fish at ∼6 months post hatch were used and 68 fish at ∼4 months post hatch were used in trial 2. Before the experiments, all fish were housed in groups of ∼50 fish in aquariums at the animal facilities at IMR, Bergen, Norway, and were fed a standard feed for zebrafish (Gemma micro 500, Skretting United States, Tooele, UT, United States). Although the trials were performed approximately 2 years apart, the conditions prior to the trials were similar; the same animal facilities and feeds were used, and stable environmental conditions were recorded throughout this whole period. Trial 1 was performed between 19th of October and 19th of November 2015, while trial 2 was performed between 5th and 20th of February 2018. All work including use of isotopes was conducted in the isotope laboratory at IMR, Bergen, Norway, where safety protocols approved by the Norwegian Radiation Protection Authority (Meldenr. M3204-12) were used. Each morning, the fish to be used that same day were transferred to the isotope laboratory and housed individually in 500 mL glass beakers filled with 200 mL preheated system water from the animal facilities. Water temperature was kept stable at 28.5°C by keeping the beakers in a water bath with temperature regulation. The fish were left to acclimatize for ∼1 h and then fed one meal, ad libitum, of their randomly allocated experimental feed and left for feeding and digestion during for 30 min. Fish were then removed from the feeding beaker, flushed with system water to remove possible traces of feed on the fish, and put into a fresh beaker with no feed (500 mL glass beaker with 300 mL system water). Twenty-four hours after the transfer, the fish were euthanized by adding tricaine methanesulfonate (TMS) from a stock solution directly to the beaker (final concentration 1 g TMS L^-1). All treatments were performed each day, in randomized order and position in the water baths. The experimental procedure was conducted according to the guidelines of the Norwegian State Commission for Laboratory Animals (application number: 15/166840).

Euthanized fish were quickly transferred to ice to slow down metabolic processes, and their weight and length measured. In trial 1, the fish were dissected to remove intestine and liver (including gallbladder). The presence of intestinal content was noted, when visible. In trial 2, only whole fish were sampled and no dissections performed. In both trials, females identified during sampling were discarded and the respective treatment repeated. Each tissue sample (intestine, liver, carcass or whole fish) was put into individual glass scintillation tubes (20 mL) and received 1–2 mL of Soluene-350 (Perkin Elmer, MA, United States). The beakers were kept at ∼60°C for 4–5 h, to accelerate the breakdown of the tissues. The samples were left to dissolve overnight, or until completely liquefied, before bleaching with 30% hydrogen peroxide (2 × 100 μL) at 50–60°C for 30 min to complete decolorization. After the removal of the fish from the beakers, the water in each beaker was homogenized with a Polytron PT-MR2100 (Kinematica AG Switzerland) for 2 min to ensure fecal matter or precipitates in the water were homogenously distributed. The exact volume of water was noted and triplicate water samples (10 mL × 3) were taken from the beakers while continuously stirring the water. Both water and tissue samples then received 10 mL Ultima Gold (Perkin Elmer, MA, United States) and were left overnight to adapt to light and temperature, according to manufacturer instructions. Radioactivity was measured (2 × 10 min) in each sample with a TRI-CARB 2000CA scintillation counter (United Technologies Packard, United Kingdom).

The results in disintegrations per minute (dpm) from the scintillation counter (average dpm value of two counts of each sample) was converted to Bq for further analysis and presentation (1 Bq = 60 dpm/1000 dpm = 16.7 Bq). The value for the total fish of trial 1 was calculated as:

Total system Bq was used as an estimation of total ingested [4-¹⁴C] as well as total feed intake and was in both trials was calculated as:

Based on scintillation counts of the feeds, it was determined that fish with a Bq_totalsystem value lower than ∼33 Bq (2000 dpm) ate less than one third of an average grain of feed (data not presented). This was therefore set as a lower limit in order to separate fish that did not eat from the fish that did eat. All data is presented as mean ± standard error of the mean (SEM) if not otherwise stated.

All statistical analyses were done using the free software environment R (R Development Core Team, 2017). Effects of dietary treatments (categorical) on the excretion of cholesterol and tissue distribution were analyzed with linear mixed effects models using the nlme package (Pinheiro et al., 2010) (nested ANOVA, random effect: sample day) followed by Tukey’s HSD using the package multcomp (Hothorn et al., 2008). For regression analyses of effects of numerical values (dietary phytosterol concentration and ratio, phytosterol dose per fish, phytosterol dose per mg bodyweight) on cholesterol excretion and tissue distribution, nlme was also used (random effects: sample day and trial), but here the EZE groups were excluded from the statistical analyses to see only the effects of sterol dose.

The fish of trial 1 and 2 weighed 559 ± 9.7 mg and 420.4 ± 4.8 mg, respectively. There were no significant differences between the treatments in weight or condition factor within the two trials (Supplementary Tables 4, 5). Each individual fish had a mean feed intake (calculated based on Bq in total system and Bq in feeds) equal to a mean of ∼1.8 mg of feed eaten (0.3% of bodyweight) in trial 1 and ∼1.2 mg (0.3% of bodyweight) in trial 2 (Supplementary Tables 4, 5). The actual dose of phytosterols per weight unit of each individual fish was calculated and varied from 0.0009 to 0.0078‰ of bodyweight (Supplementary Tables 4, 5).

There was no effect of the dietary phytosterols on the excretion of [4-¹⁴C] cholesterol in neither trial 1 nor trial 2. The dietary effects on cholesterol excretion were statistically evaluated by comparing between the treatment groups as well as by using the actual dose of phytosterols based on amount of feed eaten and dietary analyses (dietary phytosterol concentration and ratio, phytosterol dose per fish, phytosterol dose per mg bodyweight) and no effects were found due to any of these parameters. The excretion of [4-¹⁴C] cholesterol plotted against the analyzed dietary phytosterol:cholesterol ratios is presented in Figure 1A. The ezetimibe treatment was very efficient at increasing the cholesterol excretion, more than doubling the mean excretion from 29.1 ± 1.7% (all non-EZE groups from both trials, n = 96) to 71.9 ± 2.5% (EZE groups from both trials, n = 20) (Figure 1B). The EZE groups had a significantly higher cholesterol excretion than the other treatment groups in both trials (p < 0.001, nested ANOVA). Of the [4-¹⁴C] cholesterol in the fish tissues (Bq_{total fish}) measured in trial 1, 58.5 ± 1.8% (n = 48) was still in the intestine 24 h after the meal. The liver contained 5.3 ± 1.2% and the remaining carcass 36.2 ± 1.3% (n = 48) (Figure 2). Neither the phytosterols nor the ezetimibe treatment had any effect on the tissue distribution of the dietary cholesterol (data on [4-¹⁴C] distribution in all treatment groups in trial 1 presented in Supplementary Figure 2). Out of a total of 48 fish dissected and analyzed in trial 1, only five fish had visible traces of digesta in the intestine. However, the visible presence of digesta did not have any effect on the dpm readings of the organ, likely due to a fast absorption of dietary cholesterol into the intestinal tissue, leaving only more slowly digestible nutrients in the intestinal lumen.