The experimental design and fish use protocol were approved by the Institutional Animal Care and Use Committee (IACUC) of Dartmouth College. Also, we conducted all experiments in accordance with relevant guidelines and regulations. We euthanized the fish by single cranial pithing in the nutritional feeding experiment.
Diet formulation for nutritional feeding experiment
We incorporated N. oculata defatted biomass to replace different percentages of FM and whole cell Schizochytrium sp. to replace all FO in three tilapia experimental diets for a nutritional feeding trial. These three diet formulations were based on our previous digestibility data for N. oculata defatted biomass and whole cell Schizochytrium sp.17,30,33, and a prior study showing potential to replace all FO with whole cell Schizyochytrium sp.30. We compared these three experimental diets to a reference diet (served as control diet) containing FMFO at levels found in commercial tilapia feed. All diets were iso-nitrogenous (37% crude protein) and iso-energetic (12 kJ/g). Microalgae inclusion diets used N. oculata defatted biomass to replace 33% (33NS), 66% (66NS), and 100% (100NS) of the FM and whole cell Schizochytrium sp. to replace all FO in the test diets (33NS, 66NS, 100NS). Thus N. oculata comprised 3%, 5% and 8% of the diet by weight, respectively, and Schizochytrium sp. made up 3.2% of the diet by weight. We produced the diets in accordance with our previous work17,30,36. We obtained dried Schizochytrium sp. from ALGAMAC, Aquafauna Bio-marine, Inc., Hawthorne, CA, USA; and menhaden FO from Double Liquid Feed Service, Inc., Danville, IL, USA. Qualitas Health Inc., which markets EPA-rich oil extracted from N. oculata as a human supplement39 and seeks uses for tons of under-utilized defatted biomass from its large-scale production facilities, donated the N. oculata defatted biomass. Supplementary Table S8 reports proximate compositions and amino acid profiles of N. oculata defatted biomass and Schizochytrium sp.; total fatty acid profile by percentage of the defatted biomass and Schizochytrium sp ingredients reported in Supplementary Table S9; and macromineral and trace element composition of both ingredients reported in Supplementary Table S10. The formula, proximate analysis, and amino acid profiles of four dietary treatments reported in Table 1. The fatty acid profiles reported in Supplementary Table S11 and the macrominerals and trace elements of the four experimental diets reported in Supplementary Table S7.
Experimental design and sampling to evaluate tilapia growth on N. oculata defatted biomass and Schizochytrium sp. Diets
We conducted the feeding experiment using a completely randomized design of four diets × three replicates tanks in recirculating aquaculture systems (RAS). Four hundred eighty Nile tilapia (mean initial weight 34.5 ± 2.06 g) were put into randomized groups of 40, bulk weighed, and transferred to a tank. Tilapia had been acclimated to the FMFO containing reference diet for 7 days prior to distribution. The initial stocking density remained within levels recommended to avoid physiological stress on tilapia (< 0.25 lbs/gal in 80 gallon RAS tanks). We carefully monitored water quality daily to maintain favorable conditions for tilapia across all RAS tanks and kept the water temperature at 28.7 ± 0.25 °C, pH at 7.1 ± 0.1, dissolved oxygen at 6.1 ± 0.15 mg/L, total ammonia nitrogen at 0.26 ± 0.1 mg/L, and nitrite nitrogen at 0.3 ± 0.01 mg/L17,30.
We administered feed at a rate of 8% of body weight until day 60, 6% until day 121, and 4% until day 183, with feedings performed twice per day at 09:00 and 15:30 h. We measured fish biomass monthly by randomly selecting 10 fish as a weight sample to adjust feeding rates for growth and we bulk weighed all fish every other month for sampling events (day 0, 60, 121, and 185). We withheld feed for 24 h prior to the weighing procedure to reduce handling stress on fish.
Biological sampling and tissue collection
We randomly selected and weighed 10 individual fish from the total starting stock at the beginning of the experiment, then euthanized (by single cranial pithing17, and stored fish tissues at – 20 °C for future biochemical analysis. At day 121 of the experiment, we euthanized 6 fish per tank, and 6 additional fish at day 185, the terminus of the trial. Half of the fish sampled on day 121 and day 185 were filleted, and half were kept whole and then stored at − 20 °C for further processing17,30. All samples from the initial sampling, day 121, and day 185 were freeze dried at − 20 °C, then fully homogenized. Both whole body and fillet samples were sent to New Jersey Feed Laboratory, Inc (Ewing, NJ, USA) for full proximate, energy, amino acid, and fatty acid profiles.
Analytical procedure and calculation
We quantified final weight, weight gain, weight gain percentage, FCR, SGR, PER, and survival rate for each of the dietary treatments. Each of these parameters were calculated as follows: weight gain = (final weight − initial weight/initial weight) × 100; FCR, FCR = feed intake/weight gain; protein efficiency ratio; SGR (%/day) = 100 × ln final wet weight (g) − ln initial wet weight (g))/Time (days), PER = weight gain (g)/protein fed (g); and survival rate (%) = (final number of fish/initial number of fish) × 10017,34,40.
The trace mineral content of each of the experimental diets, sampled fish fillets, and whole bodies was analyzed by the Department of Earth Science at Dartmouth College17. Each 100 mg sample was acid digested in 0.5 mL 9:1 HNO3/HCl in open vessel digestion with heating at 105 °C for 1 h. Samples were diluted to 10 mL in DI water prior to analysis. All measurements were recorded gravimetrically. Digested samples were run by ICP-MS analysis using an Agilent 7700 × with collision (He) and reaction (H2) gases. The methodology and quality control followed EPA method 6020a.
Degree of protein hydrolysis and in-vitro protein digestibility
We performed an in-vitro digestibility assessment according to the method prescribed in Yasumaru and Lemos to measure the degree of protein hydrolysis of our experimental diets in the presence of tilapia stomach crude enzyme extract and intestine crude enzyme extract41. A 50 g sample from each of the four diets was ground via mortar and pestle until all materials could fit through a 0.5 mm food sieve. We allotted 80 mg by protein basis of each diet with 25 mL DI water in a 50 mL reaction vessel immersed in a water bath held at 25 °C. The reaction mixture, containing diet and DI water, was adjusted to pH 2.0 with 0.1 M HCl using a Hannah instrument HI-901C1 potentiometric auto titrator, set to dose 0.3 mL HCl every 2 min for 30 min until pH equilibrium was reached. After equilibrium, we introduced 200 µL stomach crude enzyme extract prepared according to Yasumaru and Lemos with storage solution modifications sourced from Chaijaroen and Thongruang41,42. After crude enzyme extract introduction, we made minor pH changes adding 0.1 M HCl or 0.01 M NaOH by hand when necessary. Once we introduced the crude enzyme extract, we initiated a predetermined program on the auto titrator to dose 0.025–0.075 mL in proportion to the change in pH measured. This program dosed accordingly every 3-min interval to keep the pH at 2.0 for 1 h. The program was paused, when necessary, to prevent over adjusting the solution during the titration. After the 1-h stomach digestion period, we recorded the total volume dosed. We then adjusted the reaction mixture pH to 8.0, using 0.1 M NaOH, and allowed the auto titrator to dose 0.025 mL 0.1 M NaOH for approximately 1 h to allow the mixture to reach equilibrium. Once pH equilibrium was reached, we introduced 250 µL intestinal crude enzyme extract, prepared in the same way as the stomach crude enzyme extract. Minor adjustments to pH were made by hand using 0.01 M NaOH or 0.1 M HCl. Then we initiated the auto titrator method to dose 0.01–0.025 mL 0.1 M NaOH proportional to the measured change in pH, in order to hold the pH at 8.0 for 1 h, and recorded the total volume dosed. All diets were run in triplicate41,43. We quantified the degree of protein hydrolysis in the stomach using the following equation:
$$DH = left[ {frac{V times N}{E}} right] times left( frac{1}{P} right) times F_{pH} times 100% ,$$
(1)
where DH is the degree of hydrolysis, V is the volume of the acid consumed (mL), N is the normality of the acid (H+ available for release × Molarity), E is the mass of the substrate protein (g), P is the number of peptide bonds cleaved (mol g protein−1) and when amino acid composition is unknown, (8.0), and FpH is the correction factor for pH 2.0 at 25 °C (1.08).
We quantified the degree of protein hydrolysis in the intestine using the following equation:
$$DH = B times Nb times left( frac{1}{a} right) times left( frac{1}{MP} right) times left( {frac{1}{{H_{tot} }}} right) times 100% ,$$
(2)
where B is the volume of alkali consumed (mL), Nb is the normality of the alkali (alkali groups × Molarity), a is the average degree of dissociation of the a-NH2 groups (1/a = 1.50 for pH 8.0 at 25 °C), MP is the mass of substrate protein (g), and Htot is the total number of peptide bonds in the protein substrate [7.6–9.2 meqv g protein−1] according to the source of protein44.
After calculating the degree of protein hydrolysis, we determined the in vitro protein digestibility using a prediction equation model as reported by Yasumaru and Lemos and Tibbets41,43. The degree of protein hydrolysis was used as input in the following equation to determine in vitro protein digestibility, IPD = (3.5093DH + 70.248).
Economic analysis of fish-free feed formulated with microalgae blends
We obtained commodity and market prices for the formulated feed ingredients from a variety of sources (Supplementary Tables S5 and S12). We conducted non-parametric bootstraps in RSTUDIO (v.1.2.5033) based on 10,000 replicates using the adjusted bootstrap percentile method to estimate the median and 95% confidence intervals.
We conducted a hedonic analysis in RSTUDIO to estimate the price of defatted N. oculata meal and whole cell Schizochytrium sp. The general methodology of hedonic analysis is described in Maisashvili et al.45. We used mixed-effects linear models using maximum likelihood methods46,47.
Following Maisashvili et al., we selected crude protein, ether extract, methionine, and lysine as the key input variables in our defatted N. oculata meal model45. We used the following regression formula:
$${varvec{y}}_{t} = beta_{0} + b_{{0,CP_{t} }} + b_{{0,EE_{t} }} + beta_{1} cdot {varvec{CP}}^{2} + beta_{2} cdot {varvec{Met}}^{2} + beta_{3} cdot {varvec{Lys}}^{2} + b_{{1_{t} }} cdot {varvec{CP}} + left( {beta_{4} + b_{{2_{t} }} } right) cdot {varvec{EE}} + varepsilon ,$$
(3)
where yt is the vector of feed ingredient prices observed at time t, CP is a vector of independent variables reflecting the crude protein content of the corresponding feed ingredients, Met is a vector of independent variables reflecting the methionine content of the corresponding feed ingredients, Lys is a vector of independent variables reflecting the lysine content of the corresponding feed ingredients, EE is a vector of independent variables reflecting the ether extract content of the corresponding feed ingredients, β0 is the fixed-effect intercept, β1 is the fixed-effect coefficient of CP2, β2 is the fixed-effect coefficient of Met2, β3 is the fixed-effect coefficient of Lys2, β4 is the fixed-effect coefficient of EE, b0,CP is the random-effect intercept of CP at time t, b0,EE is the random-effect intercept of EE at time t, b1 is the random-effect coefficient of CP at time t, b2 is the random-effect coefficient of EE at time t, ε is the residual error, and t is the time period (2010–2019).
We selected the top fatty acids present in both the commodity oils (vegetable and fish) and in Schizochytrium sp. that did not require an extrapolation. Thus, we used the following regression formula:
$${varvec{y}}_{t} = beta_{0} + b_{{0,14:0_{t} }} + b_{{0,16:0_{t} }} + beta_{1} cdot {user2{20:5}}{mathbf{n}}{ – }{user2{3}}^{2} + beta_{2} cdot {user2{14:0}}^{2} + beta_{3} cdot {user2{16:1}}{varvec{n}}{ – }{user2{7}}^{2} + left( {beta_{4} + b_{{1_{t} }} } right) cdot {user2{14:0}} + left( {beta_{5} + b_{{2_{t} }} } right) cdot {user2{16:0}} + varepsilon ,$$
(4)
where yt is the vector of oil ingredient prices observed at time t, 20:5n-3 is a vector of independent variables reflecting the EPA content of the corresponding oil ingredients, 14:0 is a vector of independent variables reflecting the myristic acid content of the corresponding oil ingredients, 16:1n-7 is a vector of independent variables reflecting the palmitoleic acid content of the corresponding oil ingredients, 16:0 is a vector of independent variables reflecting the palmitic acid content of the corresponding oil ingredients, β0 is the fixed-effect intercept, β1 is the fixed-effect coefficient of 20:5n-32, β2 is the fixed-effect coefficient of 14:02, β3 is the fixed-effect coefficient of 16:1n-72, β4 is the fixed-effect coefficient of 14:0, β5 is the fixed-effect coefficient of 16:0, b0,14:0 is the random-effect intercept of 14:0 at time t, b0,16:0 is the random-effect intercept of 16:0 at time t, b1 is the random-effect coefficient of 14:0 at time t, b2 is the random-effect coefficient of 16:0 at time t, ε is the residual error, and t is the time period (2010–2019).
As inputs to Eqs. (3) and (4), we used the mean annual prices for 12 meal ingredients and 7 oil ingredients from January 2010 to December 2019 (see Supplementary Table S12 for details about the commodities and data sources). Although some studies have used shorter time horizons for their hedonic models (e.g. 2 years)48, we followed other studies that used longer time horizons (e.g. 10 years) in their hedonic models49 and economic analysis of agricultural commodities to capture variability50. We incorporated a freight component to calculate the costs to bring these commodities to the Port of Shanghai, China. To account for the multi-modal components of the freight costs of U.S. commodities, we applied modal transport shares (e.g. rail, truck, barge) of grain commodities (e.g. corn, wheat, soybeans, sorghum, and barley) to the distances between the grain production sites and U.S. ports (see Supplementary Table S13 and Supplementary Methods for further details). We used a shipping route distance calculator to estimate the international shipping distances (Supplementary Table S14). We obtained the nutritional composition of the feed commodities from Archer Daniel Midlands and Feedinamics (Supplementary Table S15). We obtained the fatty acid profiles of the oils used in the feed from the literature (Supplementary Table S16). For the terrestrial-plant-based oils, we used the fatty acid values reported in Dubois et al.51. For FO, we used the fatty acid values reported in Sarker et al.30. We scaled the vectors of independent variables (Supplementary Tables S15 and S16) with the parameters provided in Supplementary Tables S17 and S18, for defatted N. oculata and whole cell Schizochytrium sp., respectively. We assessed the goodness of fit using graphical methods and diagnostic tests (see Supplementary Methods, Supplementary Tables S19 and S20, and Supplementary Figs. S2–S7 for further details).
We estimated the price of defatted N. oculata meal with Eq. (3), the scaled parameters (Supplementary Table S21), the fixed-effect coefficients (Supplementary Table S22), and the random-effect coefficients (Supplementary Table S23). We estimated the price of whole cell Schizochytrium sp. with Eq. (4), the scaled parameters (Supplementary Table S24), the fixed-effect coefficients (Supplementary Table S25), and the random-effect coefficients (Supplementary Table S26). To convert the estimated price of Schizochytrium sp. oil to whole cell Schizochytrium sp., we multiplied the price by the fraction of lipids in Schizochytrium (0.54).
We calculated the costs of all ingredients of formulated reference feed and experimental feeds (which combined N. oculata defatted biomass with Schizochytrium sp.) to determine the diet costs in USD per kg (Supplementary Table S27). The price of each diet was determined by multiplying the respective contributions of each feed ingredient by their respective costs per kg and summing the values obtained for all of the ingredients in each of the formulated diets. Finally, we estimated the production cost of tilapia ($/kg fish) via ECR to compare among the four experimental tilapia feeds (which combined defatted biomass with Schizochytrium sp.). We estimated fish production cost as ECR using the equation of Piedecausa52:
$$ECR left( {frac{$ }{{{text{kg}},fish}}} right) = FCR left( {frac{{{text{kg}}, diet, fed}}{{{text{kg}} ,weight ,gain}}} right) times price ,of, diet left( {frac{USD$ }{{{text{kg}} ,diet}}} right),$$
(5)
where ECR is the economic conversion ratio, and FCR is the feed conversion ratio.
Statistical analysis
Statistical analysis (ANOVA) was performed according to Sarker et al.17 to determine the significant differences in proximate and amino acid content, fatty acid profile, final weight, weight gain, weight gain percentage, in vitro protein digestibility, FCR, SGR, PER, survival rate, and ECR for each of the treatments. When significant differences were found, we compared the treatment means using Tukey’s test of multiple comparisons (posthoc), with a 95% confidence interval. The IBM Statistical Package for the Social Sciences (SPSS) program for Windows (v. 21.0, Armonk, NY, USA) was used for all statistical methods.
Data and code availability
The datasets and RSTUDIO files used in the economic analysis including the hedonic regression analyses (used to estimate the price of defatted N. oculata meal and whole cell Schizochytrium), bootstrap confidence intervals of feed ingredient prices, and the ECR for Fig. 2 are available at the following link: https://doi.org/10.6071/M3VD5V.
Source: Ecology - nature.com