Feed Intake as an Estimation of Attractability in Pacific White Shrimp Litopenaeus vannamei

Abstract

A simple bioassay that quantifies feed intake as an estimation of relative attractability of feeds containing different ingredients in the Pacific white shrimp Litopenaeus vannamei is described. Fish meal (FM), fish protein hydrolysate (FPH), squid meal (SqM) and casein (CN) were assessed at the same dietary level for their relative influence on feed intake rates of Litopenaeus vannamei. A bland diet containing 92% whole wheat grain meal, 6% diatomaceous earth and 2% alginate with a known low attractability was used as the standard control or base diet. Ingredients were added to the bland base control diet at a level of 3% as fed. Shrimp were stocked into 80 L glass tanks (n= 20 per tank) in a recirculating aquaculture system. Tanks were randomly assigned to one of five diet treatments (3tanks/treatment). Experiments measuring the attractability of each feed were conducted twice daily at 0900 hours and 1330 hours over a five day period. For each experiment, 40 feed pellets (ca. 1 g) corresponding to the assigned treatment were provided to each tank. To calculate the rate of feed intake, pellets remaining in each tank were counted at six minute intervals for a seventy-two minute period. Differences in rate of feed intake among diets were evaluated using Cox Regression Analysis. This attractability assay required only small amounts of ingredients and incorporated ingredients into a bland feed, which significantly reduces the influence from other ingredients or compound in the pellets. All of the test protein ingredients, especially SqM, in the feeds significantly increased the feed consumption rate. The diet containing SqM was consumed at a significantly higher rate than those containing casein and FM but not FPH. FPH and CN containing diets were not significantly different but consumed at a higher rate than the diet containing FM. Results of these trials indicate that the presence of certain ingredients can increase feed intake, thereby increasing nutrient availability of the diets. This reported method to determine consumption of diets containing certain ingredients can be considered as a valid method to estimate attractability.

1. INTRODUCTION

The commercial production of farmed shrimp has been expanding steadily. World aquaculture production of Litopenaeus vannamei was over 4.5 million tons in 2018, an increase of 3 to 5 percent over 2017 (FAO 2018). As shrimp aquaculture is expected to continue to increase in coming years, the animal feed industry will need consistent and predictable supplies of ingredients.

About 69% of world fish meal was used for aquaculture in 2016 (IFFO Fishmeal and Fish Oil Statistical Yearbook 2017), and commercial feeds for aquaculture continue to use a very large amount of fish meal contributing to the world fisheries being at or near maximum sustainable yield for many marine fishes. Commercial shrimp feeds typically contain 10 – 30% of fish meal as a high quality protein source (Tacon and Barg, 1998; Venero et al., 2008) and for attractability (Suresh and Nates, 2011). Increasing economic and ecological concerns regarding the use of fish meal in diets for shrimp have led research institutions and aquaculture industries to search for suitable and less expensive alternative protein and attractability sources with a sufficient and predictable supply to serve as a fish meal replacement.

Successful replacement of fish meal in shrimp feeds requires an evaluation of attractability of alternative protein sources. Shrimp diets with a comparatively high level of vegetable protein sources are often linked with reduced performance and low growth rate (Colvin and Brand, 1977; Fenucci et al., 1982; Lim and Dominy, 1995) due to decreased feed intake, resulting from lower food attractability and palatability. Shrimp completely depend on chemosensory systems to detect, locate and ingest food (Derby and Sorensen, 2008; Lim et al., 1997). Formulated shrimp diets must be chemically attractive to aid in rapid location and complete ingestion as soon as possible. Enhanced feeding behaviors may improve feed conversion and growth (Carr, 1988). Identifying the relative attractability of different alternative protein sources will assist the identification of an alternate protein source for replacement of part or all of the fish meal in commercial diets. Previous bioassays used to evaluate ingredient attractability in shrimp diets mainly focused on measuring arousal responses and shrimp orientation or movement (Lim et al., 1997; Scarfe et al., 1985). Only a few studies have assessed ingredient attractability by measuring actual feed intake per unit time (Fox et al., 2004; Sanchez et al., 2005; Walker et al., 2005). From a practical perspective, positive effects of specific ingredients on rate of feed intake could both efficiently and directly indicate the feeding enhancement property of the test ingredients. In this study, feed intake as the response variable was directly measured after each feeding as the response variable.

The objectives of this study were to develop a simple, practical attractability bioassay combined with an appropriate statistical method to compare and quantify feed ingredient attractability for L. vannamei in culture.

2. MATERIALS AND METHODS

2.1. Experimental animals and systems

Litopenaeus vannamei seedstock (ca. 30 day-old post larvae) were obtained from Claude Peteet Mariculture Center, Gulf Shores, AL and transported to the University of Alabama at Birmingham. Prior to commencement of the trial, shrimp were held in aerated holding tanks and proffered a maintenance feed of stage 1 Artemia salina nauplii (INVE) or commercial feed (ZBI Raceway 40–9, Ziegler Bros., Inc.,) to attain the initial starting weight of 3 g.

The experiment was conducted using a 4000 L recirculating aquaculture system (RAS) with artificial sea water (Crystal Sea, Marine Enterprises International, Baltimore, Maryland, USA) passing through a Polygeyser DF-3 biological filter (Aquaculture Systems Technologies, LLC, New Orleans, LA, USA), a SMART high-output 100 W UV sterilizer (Lifegard® Aquatics, Cerritos, CA, USA), and a TF500 double-venturi protein skimmer (Top Fathom, Hudsonville, MI, USA) used for foam fractionation. The system consisted of individual 80 L glass tanks (Fig. 1), with each having a bottom surface area of 0.18 m² and water depth maintained at 20.0 cm. All tanks received flow-through artificial sea water continuously with a water turnover rate of approximately 200% per hour in each tank. Municipal tap water was filtered through 5 μm sediment filter, followed by charcoal, reverse osmosis, and a cation/anion exchange resin (Kent Marine, Franklin, WI) prior to the addition of sea salt to obtain a final salinity of 28 ppt for the RAS. Tanks were covered with a white plastic grating to reduce environmental disturbances to the shrimp and loss (escape) of shrimp. Throughout the course of the experiment, animals were maintained on a 12-hour light: 12-hour dark photo-period.

Fig. 1.

Glass tank used for evaluation of attractability of feeds.

Groups of 20 shrimp (initial individual weight ca. 3.00 ± 0.15 g) were stocked into individual 80 L glass tanks with a stocking density of 118/m² or 590/m³ per tank. Tanks were assigned randomly to one of the five treatments, three replicates per treatment. Shrimp mortalities were replaced during a 48-h period post stocking.

Water quality was evaluated daily (0830 hours) and maintained at optimal levels for shrimp (Van Wyk and Scarpa, 1999) during the 5-day experimental period. Water temperature, salinity (Yellow Springs Instruments YSI-30,) and dissolved oxygen (Yellow Springs Instruments YSI-95) were measured daily (0900 hours) inside the tanks (four randomly picked tanks were tested every day). Samples of water were collected from the water sump and inside the tanks to determine total ammonia nitrogen, nitrite, nitrate, pH and alkalinity. All the analyses were conducted using saltwater test kits from Aqua Pharmaceuticals, LLC (Malvern, PA, USA) and La Motte Company (Chestertown, MD, USA). Water temperature, salinity, and dissolved oxygen were held constant (mean ± SD: 28.0 ± 0.1°C, 28.1 ± 0.4 ppt, and 5.6 ± 0.2 mg/L, respectively).

2.2. Experimental diets

Five diets were formulated and used in the assessment of attractability of test ingredients (Table 1). Crude protein content and source of the test ingredients are listed in Table 2. A bland control diet was formulated with 92% wheat grain, 2% alginate and 6% diatomaceous earth. Treatment diets were prepared by incorporating one of the four test animal protein sources [fish meal (FM), fish protein hydrolysate (FPH), squid meal (SqM) or casein (CN)] into the bland control diet. The animal protein sources were incorporated into the treatment diets at a 3% level at the expense of diatomaceous earth.

Table 1.

Ingredient composition and calculated nutrient values of diets used to assess attractability of animal protein source ingredients (as fed).

	Experimental Diets
Ingredient (g 100 g−1)	Control	Fish meal	Fish protein hydrolysate	Squid meal	Casein
Wheat grain	92.00	92.00	92.00	92.00	92.00
Alginate	2.00	2.00	2.00	2.00	2.00
Diatomaceous earth	6.00	3.00	3.00	3.00	3.00
Fish meal		3.00
Fish protein hydrolysate			3.00
Squid meal				3.00
Casein					3.00
Calculated Nutrient Values (%)a
Crude protein	9.10	11.04	11.58	11.62	11.80
Crude fat	1.02	1.30	1.32	1.20	1.03
Crude fiber	0.65	0.65	0.65	0.70	0.69
Ash	6.53	4.10	3.70	3.70	3.54

^aNutrient values on an "As Fed" basis assuming a final dietary moisture content of 10%.

Table 2.

Crude protein content (as % dry weight) and source of test ingredients used in this study.

Test Ingredient	Ingredient protein level (%)	Origin, supplier	Analysis determined by
Wheat grain	14.20%	Zeigler	Midwest Labs
Fish meal	63.45%	Zeigler	Omega Protein
Fish protein hydrolysate	81.30%	Scoular	Midwest Labs
Squid meal	82.50%	Scoular	Midwest Labs
Casein	88.30%	MP Biomedical	Midwest Labs

All test diets were produced by cold extrusion at a room temperature of 21°C. All dry ingredients (as fed) were mixed with a PK twin shell® blender (Patterson-Kelley Co., East Stroudsburg, PA) for 10 minutes and transferred to a Hobart stand mixer (Model A-200, Hobart Corporation, Troy, OH) and blended for 40 minutes. Water was added, and the mixture was blended for 10 minutes to a mash-like consistency. The diets were extruded using a meat grinder attachment (Model A-200, Hobart Corporation, Troy, OH) fitted with a 3 mm die. Diet strands were separated and dried on wire trays in a forced air oven (35°C) for 48 hours. Final moisture content of all diets was 8–10% as fed. The dried strands were hand cut into approximate uniform sizes of ca. 3 mm diameter × 3 mm length feed pellets (shaped as a cylinder) and stored individually in air-tight storage bags at 4°C until used. All diets had a moisture level of 8% to 10% and very similar physical characteristics including water stability of a minimum of four hours.

2.3. Experimental protocol

In this experiment, relative attractability of the test ingredients to L. vannamei was estimated with feed intake of the control bland diet and experimental diets over a defined time period.

Experiments measuring the relative attractability of each experimental feed were conducted twice daily at 0900 hours and 1330 hours. A circular black ceramic plate (23 cm top diameter × 15 cm bottom diameter × 2 cm width) (Fig. 2) was placed on the bottom (front) with the edge of the plate touching the front of each tank thirty minutes (0830 hours and 1300 hours) prior to adding test diets to the tanks. A PVC drop pipe of 1.5” diameter was inserted into the tank with one end above the water and the other in the center of the top of the plate. All feed pellets were hand cut to ca.3 mm x 3 mm, and then counted and weighed to ensure uniform weight per tank. For each tank, exactly 40 pellets (ca. 1 g) corresponding to a specific dietary treatment were deposited into the drop pipe in timed sequence (every 30 seconds) by tank number order onto the ceramic plate. Feed pellets were allowed to fall to the feed plate before the drop pipe was removed. Shrimp immediately began feeding on the pellets and the pellets remaining on the plate were counted and recorded at six minute intervals for seventy-two minutes. After each test period, any uneaten leftover pellets remaining in the tanks were removed by siphoning. Following the last test period for each day, feeding plates were removed from tanks and cleaned. All shrimp were then fed a proportioned high nutrient density ration greater than satiation (recommended by Texas A&M University) of commercial feed (ZBI Raceway 40–9, Ziegler Bros., Inc.,) for approximately 2.5 hours with an auto-feeder (Fish Mate F14, Aquarium Fish Feeder). Introduction of this new food ceased at 1700 hours. All feces, molts and uneaten feed were removed from each tank by siphoning at 1730 hours. All tanks were siphoned again at 0800 hours the following morning to completely remove feces and molts from the previous day. The experimental procedures outlined above were repeated daily for a five-day period. The shrimp increased in weight and had a survival of 100% during the five-day evaluate period indicating that there was no deleterious effect of the bland diet on the shrimp.

Fig. 2.

Plate used for evaluation of attractability of feeds.

For each dietary treatment, the mean number of feed pellets remaining on the plates of replicate tanks was determined for each six-minute time interval to evaluate attractability of the diets.

2.4. Statistical Analysis

To compare pellet attractability among diets, time-to-event analyses with Cox proportional hazards regression and Kaplan-Meier survival curve analysis was performed. Statistical analyses and figures were produced with R Statistical Software (R Core Team, 2017, v3.4.3) and use of the following packages in R: lme4, splines, survival, survminer, flexsurv, formatR, ggfortify, and ggplot2 packages (Bates et al., 2015; Jackson, 2016; Kassambara et al., 2017; Tang et al., 2016; Therneau, 2015; Wickham, 2016; Xie, 2012, respectively).

Data for each treatment for each experiment conducted over five days of the trial were combined for statistical analyses. Observations with increasing pellet counts were first removed to meet model assumptions as they were assumed to be counting errors. The response variable was the count of pellets remaining at each time point.

Pre-specified subgroup analyses were used to evaluate meaningful comparisons among diets. Kaplan Meier plots and Cox Hazard regression analyses were performed separately for each subgroup. Each model also included batch effects (referring to separate experiments) and time of day as covariates. The proportional hazard assumption was tested on the basis of Schoenfeld residuals and was met for all models.

The calculated hazard ratios (95% CI) represent the relative likelihood of pellet consumption compared to a reference diet (intercept). For statistically significant (p< 0.05) omnibus models, the type 1 error rate was controlled by applying Bonferroni corrections to the nested pairwise comparisons (p=0.005). While less conservative methods to adjust for multiple comparisons exist, the large number of observations for our data allowed more than adequate power to detect differences among diets with a Bonferroni-adjusted level of significance.

3. RESULTS

Consistent throughout the five-day feeding trial, shrimp immediately began feeding on the pellets after each proffering and pellets were completely consumed in most of the tanks after 72 minutes. Pellets of the bland and all experimental diets were intact at the end of the 72 minute evaluation period.

3.1. Analysis of Time of Day (AM vs PM) as an independent predictor of attractability

A comparison of the attractability data from AM and PM cohorts over the 5-day feeding trial showed that AM-fed animals consumed significantly more feed per unit time (p< 2e-16) than those fed PM (Table 3). However, the results of pairwise comparisons among diets showed that the differences among diets were similar among the AM and PM cohorts

Table 3.

Cox regression analysis matrix for comparison. Analysis of time of day (AM vs PM) as an independent predictor of attractability (all diets are included in this model).

	Attractability relative to AM	Lower 95% CI	Upper 95% CI	p value
Time Day PM	0.668	0.64	0.69	< 2e–16*

CI, Confidence interval

^*Represents a significant difference from the intercept (p < 0.005). With the Bonferroni adjustment for multiple comparisons, the adjusted α for each pairwise comparison is 0.005.

3.2. Assessment of attractability of test diets

Data from AM and PM time cohorts over the 5-day trail were combined for analyses. Data representing mean consumption of pellets (from 3 replicate tanks) for each six-minute time interval are graphically represented with Kaplan-Meier reduction curves in both linear (Fig. 3 a) and log scale (Fig. 3 b). Table 4 compares the likelihood of consumption of a pellet and probability of the significance of the comparison among dietary treatments. Diets containing fish meal (FM), fish protein hydrolysate (FPH), squid meal (SqM) and casein (CN) had a significantly (p< 0.005) higher likelihood of being consumed (1.275, 1.465, 1.617 and 1.384 times more attractable, respectively) relative to the bland control diet. The percent difference in attractiveness between diets containing FM, FPH, SqM and CN was 27.5%, 46.5%, 61.7% and 38.4% respectively from the bland control diet. This indicates that all four test ingredients contain some compounds (e.g. nucleotides, organic acids, amino acids, etc.) which improved attractability of the experimental diets.

Fig. 3.

Kaplan-Meir plots displaying the feed consumption rates for each diet. (a) The raw percentage of pellets remaining vs. time. (b) The log-transformed percentage of pellets vs. time. Each bar represents the mean of 30 observations.

Table 4.

Cox Regression Analysis Matrix for Comparison

As intercept		Control	FM	FPH	SqM	CN
Control	Cox Hazard Ratio		1.275	1.465	1.617	1.384
	95% CI		(1.16, 1.41)	(1.33, 1.62)	(1.47, 1.78)	(1.26, 1.53)
	p value		1.34e–06*	1.39e–14*	< 2e–16*	5.82e–11*
FM	Cox Hazard Ratio	0.784		1.149	1.268	1.086
	95% CI	(0.71, 0.87)		(1.05, 1.26)	(1.16, 1.40)	(0.99, 1.19)
	p value	1.34e–06*		0.004	4.45e–07*	0.085
FPH	Cox Hazard Ratio	0.682	0.870		1.103	0.945
	95% CI	(0.62, 0.75)	(0.79, 0.96)		(1.00, 1.21)	(0.86, 1.04)
	p value	1.39e–14*	0.004		0.034	0.226
SqM	Cox Hazard Ratio	0.619	0.789	0.906		0.856
	95% CI	(0.56, 0.68)	(0.72, 0.86)	(0.83, 0.99)		(0.78, 0.94)
	p value	< 2e–16*	4.45e–07*	0.034		0.0008*
CN	Cox Hazard Ratio	0.722	0.920	1.059	1.168
	95% CI	(0.66, 0.80)	(0.84, 1.01)	(0.97, 1.16)	(1.07, 1.28)
	p value	5.82e–11*	0.085	0.226	0.0008*

FM, Fish Meal; FPH, Fish protein hydrolysate; SqM, Squid Meal; CN, Casein.

95% CI, 95% Confidence interval (Lower, Upper)

^*Represents a significant difference from the intercept (p< 0.005). With the Bonferroni adjustment for multiple comparisons, the adjusted α for each pairwise comparison is 0.005.

Among diets with animal protein, diet containing SqM was consumed at a significantly higher rate (p< 0.005) than diets containing FM and CN, but not FPH (p> 0.005). While the consumption rates between diets containing FPH and CN were similar, both were higher than FM. However, the consumption rate of the experimental diet containing FM was significantly higher than that of the bland control diet.

4. DISCUSSION

Lee and Meyers (1997) proposed that consumption of food consisted of several classes of chemical stimuli relating to detection, attractant, incident, stimulation and finally continuation of consumption. Derby and Sorensen (2008) stated that shrimp depend on chemosensory systems to identify, locate and ingest food. Previous chemosensory bioassays have focused on identification of the response such as antenullar flicking (Ache, 1988) and food location behaviors such as movements toward the source of an odor (Mackie and Shelton, 1972; Murai et al., 1981; Carr et al., 1984; Harpaz et al., 1987; Zimmer-Faust et al., 1989). Thus, the detection of chemical stimuli by shrimp does not mean it is attracted to it nor does it imply that once attracted to a feed, the shrimp will consume it. By using feed consumption to evaluate diet attractability, all levels of a feeding response that involve a shrimp’s chemosensory systems were collectively measured. This method of analysis could directly and comprehensively reflect the relative attractability of each test diet.

Ingredients of animal origin, especially aquatic organisms, are feeding attractants in shrimp (Smith et al., 2005) because they are rich in small soluble chemical compounds such as certain amino acids, nucleotides and some organic acids. Each of these has been identified as a feeding stimulant and palatability enhancer. Many studies have reported that fish meal, fish protein hydrolysate, squid meal and casein serve as attractants and improve feed intake of shrimp when incorporated into the feed at optimal levels (Smith et al., 2005; Nunes et al., 2006; Grey et al., 2009; Bankefors et al., 2011; Tantikitti, 2014; Montoya-Martínez et al., 2018). Under the experimental conditions in our study, feed intake rate significantly increased when 3% of each test protein ingredients (FM, FPH, SqM or CN) was added to the bland control feed. This result is consistent and in reasonable agreement with the results reported in previously cited attractability studies. Thus, the use of feed intake as an estimation of relative attractability in L. vannamei may be considered to be a valid methodology.

A bland diet (defined as a diet formulated to have minimal attractability) was utilized in our study as a control. By incorporating each ingredient into the bland diet, the shrimp’s response to each ingredient was able to be separately and independently measured. An advantage of this method is that a comparatively small amount of ingredient can be used to test the response of a large number of individuals. Therefore, rapid and inexpensive testing of a wide range of potential protein ingredients, or other phagostimulants, is possible. Unlike feed tests incorporate an ingredient or stimuli into a complete feed composed of a number of other ingredients (Smith et al., 2005; Suresh et al., 2011), our assay incorporated ingredients into a bland feed, which minimized influences from other feed ingredients that may contain compounds that could potentially confound the feeding response.

This study demonstrates that L. vannamei are capable of discriminating between a bland diet and experimental diets containing ingredients with high protein levels (63 – 88% as fed). The ingredient proteins used in this study have previously been reported as attractants for crustaceans (Akiyama et al., 1989; Smith et al., 2005; Nunes et al., 2006; Grey et al., 2009; Suresh et al., 2011). The relatively lower consumption rate associated with the control bland feed demonstrates its suitability as a base control feed to evaluate test ingredients.

Different feed consumption rates were observed between shrimp fed the bland control diet and experimental diets, as well as between experimental diets containing different types of protein ingredients. The lower feed intake rate of shrimp fed bland control diet could have partially resulted from lower protein content (9.11% crude protein of the bland control diet versus 11.04 –11.80% crude protein of the experimental diets) with most of the lower attractability associated with the diet containing wheat grain as the main protein source. However, for multiple reasons detailed below, the differences in feed intake rate of shrimp fed different experimental diets was more likely due to the protein quality with higher attractability and/or other non-protein compounds (e.g. amino acids, nucleotides, etc.) of the test ingredients. Dietary protein contents and levels directly affect the sensitivity of feeding stimulation trials (Sanchez et al., 2005). In this study, to maximize sensitivity of feeding stimulation trails, 3% of tested protein ingredients were incorporated into a bland base diet in which the protein level is relatively low. The low overall protein levels in the diets ensured that the well-documented feeding stimulation/attractability of each tested protein ingredient was not masked. Furthermore, the four experimental diets contained very similar levels of dietary protein (11.04 –11.80% as fed basis) with a comparatively lower protein level (9.11% as fed basis) of the bland control diet. These properties significantly reduced the probability that the difference in feed consumption rates was due to the lower protein content of the bland control versus each of the experimental diets and between each experimental diet.

The pellet size in relation to the size of the animal may be considerably important in achieving maximum feed consumption (Sheppard et al., 2002). In this study, both the size and number of pellets were carefully considered. We conducted preliminary trials (data not reported) using the same size and batch of shrimp to estimate and determine the appropriate number, size and weight of feed pellets to proffer to shrimp in each tank. A 3 mm x 3 mm cylinder-like feed pellet ensured that shrimp were able to hold 1 or 2 pellets simultaneously and consume them. Forty pellets with a total weight of ca.1 g are countable, and ensured that after 72 minutes almost all the pellets were completely consumed.

The shape and dimensions of the feed plates could positively impact feed consumption (Grey et al., 2009). In this trial, identical circular black ceramic plates were used as the source of food in all tanks. The low side and large area of the plates permitted the shrimp to easily seek the pellet, enter the plate, and pick up the pellets.

It is possible that water flow characteristics could affect a shrimp’s ability to locate pellets. Water flow was not suspended during the 5-day trial. However, all shrimp were observed to distribute and orientate across the entire bottom surface of each tank, demonstrating that the water current for all tanks was similar and that the water flow was sufficiently low to not influence the distribution and activities of shrimp.

Differential water stability properties of the feed pellets can directly affect the rate of release of certain attractive chemical compounds and thus influence the relative attractability of test feed ingredients (Grey et al., 2009). In this trial, the size, shape, sinkability and stability of the diets used in this study are representative of the physical characteristics of a typical commercial shrimp diet. Furthermore, the moisture content of each tested diets was comparative to that of commercial diets and all pellets of the bland feed and all experimental feeds were intact at the end of the 72 minute experimental period. Similarities in the physical characteristics of our experimental diets during the 72 minute evaluation period indicate that the differences in feed consumption rate were not influenced by possible differences in these variables.

In summary, the method evaluated in this study is effective in discriminating differences in feeding response to ingredients, nutrients, or bioactive food components in shrimp diets. It can be accomplished without sophisticated and expensive equipment. Open source statistical packages are readily available to evaluate the results of these studies, and allow all consumption trends, not just endpoints, to be evaluated in a statistically powerful manner.