Rheological characterization of an in vitro model for salmonid chyme to quantify changes in feed composition

Abstract

Background:

Developments in the production of aquacultural salmonid feeds in the last 20 years have led to extruded diets with extremely low water content and a shift from mainly marine fish based ingredients towards plant content. These changes expose the industry to the vagaries of the highly dynamic plant protein market. Resulting variations in the precise composition of aqua feeds may carry unpredictable consequences for water quality, since some plant ingredients cause undesirable reductions in the mechanical stability of faeces. Dietary supplements known as binders that enhance the stability of faeces have the potential to mitigate these issues, but may also bring negative effects.

Objective and Methods:

The present study employs an in vitro model to perform the first fundamental rheological characterization of salmonid chyme, and a factorial experiment designed to investigate the impacts of the presence of rheologically active substances.

Results:

The highest mean viscosity values were measured for a treatment containing a 2:1 ratio of tara gum:xanthan gum, resulting in chyme four times more viscous than a control formulation containing the same amount of water. Shear resistance was quantified by analyses of slopes fitting the frequency sweep measurements.

Conclusions:

These data open a new statistical approach to monitoring the consequences of market-driven changes in aqua feed composition and their impacts on the nutrition, health or performance of farmed fish.

Keywords

Rheology digesta properties aquaculture viscosity binder

1. Introduction

Salmonids are the most commonly farmed fish family in European aquaculture with annual production exceeding 1.8 million metric tons and a value of about 7 billion Euro [1]. The carnivorous nature of most salmonid species is a significant challenge to feed manufacturers, especially with regard to the purity and quality of raw ingredients.

Modern European salmonid feeds comprise a mixture of plant proteins, plant oils, fish meal and fish oil. While these basic constituents have remained essentially unchanged over the last 15 years, there has been a fundamental shift in the ratios used, transforming what were predominantly marine fish based diets into mainly plant based feeds [2]. The growth of the aquaculture sector and simultaneous stagnation of forage fish catches have driven feed manufacturers to increase the proportion of alternative raw materials without compromising the quality of their products. Plant-derived ingredients are increasingly favored by virtue of availability, price development and consumer acceptability. Despite some emerging issues with plant-based feeds, such as anti-nutritional factors [3,4], modern salmonid feeds now contain only a fraction of the fish-based ingredient they once did, with the result that rainbow trout (Oncorhynchus mykiss) farms, for example, now represent an effective net production of fish protein.

But the composition of modern feeds is far from stable. Highly dynamic market forces mean that the feed industry must constantly adapt to new realities, leading to almost continual changes in feed composition.

Feed composition is a principle factor determining the physico-chemical (mechanical) properties of gastric and intestinal chyme and of faecal matter. This crucial but previously underestimated factor in aquaculture can have grave consequences not only for performance and welfare but also for farm-scale operations [5–7]. Faeces and the accumulation of resulting suspended solids are a major issue in all forms of aquaculture, contributing to the environmental impact of flow through systems and impacting on process stability and the health and welfare of stock in closed containment systems with high water reuse rates [8–10].

Decisive factors affecting the physical properties of chyme and faeces include the presence of substances that influence the intrinsic cohesiveness of the complex fluid-particle matrix. In the context of salmonids, indigestible carbohydrates such as the soluble non-starch polysaccharides (NSPs) are of considerable importance. NSPs are naturally present in a variety of plant based protein sources and may also be deliberately added to feeds for thickening purposes. The potency of NSPs as binders varies, but some can strongly affect the viscosity and elastic modulus of chyme and faeces, even at very low concentrations [5].

Indigestible structural components of chyme accumulate during the digestion process as other digestible components are absorbed, so that the highest concentrations and strongest effects tend to be expressed in the faecal cast. However it is also likely that some effects of rheologically active components will begin to be apparent much earlier, during the gastric and intestinal phases of digestion. An overview of the rheological conditions in the gastrointestinal system of salmonids is given in Fig. 1.

Fig. 1.

Schematic overview of the gastrointestinal system of rainbow trout focusing on factors influencing rheological conditions.

Apart from the anti-nutritive effects [4], the physical properties of chyme are determined mainly by the indigestible fiber fraction, which can influence a variety of different processes, from transport to digestion along the entire length of the gastrointestinal (GI) tract.

Preliminary work on the rheological behavior of fish faeces was carried out by Brinker et al. [5]. The present study uses a more fundamental approach, performing a classic rheological characterization of an artificial gastric digesta with properties closely resembling those of salmonid faeces.

The effects of time, a frequency spectrum and a linear stress ramp on the flow behavior of the model material were investigated. Special emphasis was placed on the detection of structural differences induced by changes in salmonid feed composition. The four different NSP-supplements used to develop chyme models with differing physico-chemical properties brought an applied element to the study, as some are already used in commercial feeds in order to manipulate the properties of faeces, though detailed knowledge of their impacts on digestive processes is lacking. This investigation seeks to establish hitherto unreported relationship between NSP additives and the structural properties of chyme while adding to the existing scarce knowledge of how the material properties of chyme impact on digestion.

2. Materials and methods

2.1. Raw materials

All feeds comprised an identical mixture of three main components: a basal premix (Biomar AS, Denmark); an oil mix and water. The ingredients of the basal premix are shown in Table 1. The lipid fraction was predominantly fish oil (origin Peru), with a smaller fraction of cold-pressed Frisian rapeseed oil. Crude compositions of the different diets are given in Table 1. The binder-free diet 1 was used as reference control, to which composition of all other diets differed only in the inclusion of 0.2% of four different NSPs in various combinations, as shown in Table 2. The supplemented combinations of NSPs were selected for their potential to achieve changes in chyme properties at low concentrations, based on information previously recorded in the literature [11–14].

Table 1
Dry matter concentrations of non-starch polysaccharides (NSP) in the experimental diets prior to extrusion, and cold viscosity (mPas) values provided by manufacturers based on assessment under standard conditions (Seah International, France)

(g kg⁻¹) Cold viscosity^Ψ(mPas) (individual binder) Diet 1 Diet 2 Diet 3 Diet 4 Diet 5

Guar gum (GG)^a (9500–10.000) – 2 0.67 – –

Xanthan gum (XG)^a (1200–1600) – – 1.33 0.67 1.2

Locust bean gum (LBG)^a (4500–6500) – – – – 0.8

Tara gum (TG)^a (2900) – – – 1.33 –

Final NSP concentration – 2 g kg⁻¹

(g kg⁻¹)	Cold viscosity^Ψ(mPas) (individual binder)	Diet 1	Diet 2	Diet 3	Diet 4	Diet 5
Guar gum (GG)^a	(9500–10.000)	–	2	0.67	–	–
Xanthan gum (XG)^a	(1200–1600)	–	–	1.33	0.67	1.2
Locust bean gum (LBG)^a	(4500–6500)	–	–	–	–	0.8
Tara gum (TG)^a	(2900)	–	–	–	1.33	–
Final NSP concentration		–	2 g kg⁻¹

Seah International, France.

^Ψ

(Viscosity 1% solution, Brookfield RVTDII, Spindle RV3 12 RPM, Temperature: 25°C.)

Table 2

Adjusted mean values (with their standard errors) of dynamic viscosity $η^{'}$ and elastic modulus $G^{'}$ of the time sweeps at different water contents (weight basis wt%) without and with inclusion of respective non-starch polysaccharide(s) (NSP)

Diet	NSP	Viscosity (Pa*s)^Ψ	Elastic modulus (Pa)^Ψ	Water content (wt%)
1	–	76.9^h ± 0.9	244.2^fg ± 4.5	60%
2	GG	156.0^d ± 1.0	439.8^d ± 4.9
3	$XG + GG$	221.2^c ± 1.0	728.8^c ± 4.9
4	$XG + TG$	278.0^a ± 1.0	877.7^a ± 4.9
5	$XG + LBG$	234.0^b ± 1.3	761.8^b ± 6.3
1	–	26.0^k ± 1.0	68.6ⁱ ± 4.9	62.5%
2	GG	68.4ⁱ ± 1.0	157.0^h ± 4.9
3	$XG + GG$	121.8^e ± 1.0	373.2^e ± 4.9
4	$XG + TG$	94.0^f ± 1.0	264.6^f ± 4.9
5	$XG + LBG$	82.5^g ± 1.0	237.7^g ± 4.9
1	–	9.8^l ± 1.5	24.9^j ± 7.7	65%
2	GG	21.4^k ± 1.0	39.5^j ± 4.9
3	$XG + GG$	32.5^j ± 1.0	89.7ⁱ ± 4.9
4	$XG + TG$	32.4^j ± 0.9	79.4ⁱ ± 4.5
5	$XG + LBG$	34.2^j ± 1.0	91.0ⁱ ± 4.9
		Model effects^†
Diet		***	***
Water content		***	***
Interaction		***	***

^Ψ

$Adjusted means \pm standard error (SE)$ . Values within a column not sharing a superscript letter are significantly different.

(Tukey–Kramer HSD; $p < 0.05$ ).

^†

Model effects: *** $p < 0.0001$ .

2.2. Feed production

The dry feed ingredients were mixed together for 30 minutes in a tumble mixer, then transferred to a fluid mixer and processed into a homogenous dough with the gradual addition of the oil and water over 2 min at 1000 rpm. The mixture was immediately extruded using a twin screw compounder (Collin Teach-Line Zk 25T) under conditions close to those in commercial feed manufacture facilities (barrel temperature at feed matrix: 90°C, max pressure at die 0.75 MPa).

Extruded feed was dried in an oven for 40 min at 135°C and subsequently stored under cool, airtight conditions ( $\sim 4 ° C$ ) until usage/measurement.

2.3. Preparation of model chyme

For the preparation of the model chyme, feed was crushed in a blade granulator and passed through a 50 micron sieve. Components of chyme were weighed by a precision scale and samples were prepared in distilled water by stirring gently for 20 minutes at room temperature under humid conditions. Chyme mixtures from five different diets (Table 1) were studied at three different liquid water levels (60%, 62.5%, 65%), within the range of gastric moisture levels previously reported in salmonids fed commercial dry feeds [15–18]. The consistency of the prepared chyme ranged from suspension-like to paste-like (depending on water level and treatment). Total binder concentrations were the same for all treatment diets and following dilution to approximate physiological moisture contents amounted to 0.08 weight percentage (wt%) for the samples with 60% water, and 0.075 wt% and 0.07 wt% for samples with 62.5% and 65% water respectively.

2.4. Rheology

Rheological measurements were performed on a Paar Physika MCR 301 (Anton Paar, Austria) using a parallel plate geometry, with a PP-50 measurement system (50 mm diameter, 1 mm gap, $sample size \sim 2 mL$ ). Samples were loaded carefully onto the plate to minimize structural damage, and the measuring area was sealed within a humidity chamber to prevent anomalies due to evaporation.

2.4.1. Oscillatory and creep measurements

The rheological characterization of the model chyme was performed using a three-part measuring sequence comprising an initial time sweep, a short delay of 60 s, a frequency sweep, a further rest of 120 s and then a creep test applied with a linear stress ramp.

Prior to the measurements described above, amplitude tests were carried out on samples from each treatment in order to identify the linear viscoelastic regions (LVEs) within which the oscillatory tests would be performed [19]. Temperatures were maintained using a Pelletier system at a constant of 15°C, a standard temperature for rainbow trout farming.

Time sweep tests. Time sweep tests were conducted at an angular frequency ω of 1 rad/s and a constant strain amplitude γ of 0.1% (within the predetermined LVE region) via 15 measuring points and with a total test duration of 300 s.

Frequency sweep tests. The frequency dependence was determined from 0.1 rad/s to 100 rad/s at 16 measuring points using a logarithmic scale and constant strain amplitude γ of 0.1%.

Creep test. The creep tests were performed after a 120 s resting phase. The sample was subjected to a stepwise positive stress ramp from 0 to 1 Pa and the strain was recorded as a function of time (with a duration of 1000 s). The rate of stress change was set at 0.001 Pa/s.

Analyses of rheological data. For time sweep measurements, adjusted means (least square means over all 16 measuring points) were used to detect differences between treatments.

In the case of frequency sweeps, the slopes for each distinct region within the double logarithmic plots were determined by linear fits. The slopes of both regions and the sum of the two were then analysed further to characterize differences.

The creep curves were plotted on double logarithmic scales and fitted respectively with the following equation: $\begin{matrix} (1) & γ = A σ^{a}, \end{matrix}$ where γ is deformation and σ is shear stress; fit parameters A and a were analysed for differences between treatments.

2.5. Statistical analysis

Experiments were designed by means of power analysis using the DOE tool implemented in JMP Pro 13 (SAS institute). As the present work is novel no lab data have been available. Therefore, power analyses were informed using expected mean differences and variations according to values reported in Brinker et al. [5]. The power analyses analyzed 5 means with four additional factors as applied in the final generalized linear model (GLM) of our manuscript (cf. M&M). As Brinker et al. [5] analyzed real samples, lower variations of the in vitro chyme measurements were expected. The power analyses resulted in a statistical power close to 1 which is in line with expectations as observed mean differences between diet samples are large, whereby variation is comparably small. The real data of the present paper are in line with these assumptions.

All measurements were performed at least five times (except only twice for the control of 65% and three times for the control of 60% water content). Differences in the adjusted means of viscosity and storage modulus of the time sweeps, sums of the slopes of frequency sweeps and the parameters fitting of the creep experiments were tested using t-tests and, in cases of heteroscedasticity, Welch’s test [20]. For diet-dependent analyses of the rheological data, the following linear parametric model was applied: $\begin{matrix} (2) & Y_{i j k} = μ + a_{i} + b_{j} + {(a b)}_{i j} + {[c]}_{k} + ε_{i j k l}, \end{matrix}$ where $Y_{i j k}$ is the evaluated parameter, μ is the overall mean, $a_{i}$ is the fixed factor Diet and $b_{i}$ is the fixed factor Water Content of the different diets, ${(a b)}_{i j}$ denotes the interaction between Diet and Water Content, ${[c]}_{k}$ is the random factor Measuring point and $ε_{i j k l}$ is the random residual error. Post hoc comparisons were carried out using Tukey’s HSD (honest significant difference) test [21]. Significance was accepted at $p < 0.05$ . All statistical analyses were performed using JMP Pro (SAS Institute Inc.), version 13.

Fig. 2.

Time sweep curves (means values with their standard errors) for five diet treatments at 60% water content.

3. Results

3.1. Time sweep measurements

An overview of the time sweep measurements for all diets with a moisture level of 60% is given in Fig. 2. The model chyme showed solid-like behavior (with $G^{'} > G^{″} = η^{'}$ under given conditions) in all cases. Time sweep measurements of nearly all treatments displayed similar curve progressions, with decreasing viscosity and increasing elastic modulus over time. One exception was the control diet with a water content of 65%, for which no changes in viscosity over time were observed (data not shown). The degree of temporal variation was affected by the factors “diet” and “water content” ( $p < 0.001$ ). For each diet, the mean values for dynamic viscosity $η^{'}$ and elastic modulus $G^{'}$ were clearly distinguishable between samples with 60% and 62.5% water content, indicated by highly significant differences in adjusted means (Table 2). Across the tested range of water contents, the control diet 1 showed the lowest values for viscosity and elastic modulus, followed by the guar gum treatment. For formulations with a water content of 60%, diet 4 exhibited highest viscosity and elastic modulus, followed by diet 5 and diet 3. At 62.5% moisture, however, highest values for viscosity and elastic modulus were recorded for diet 3, followed by diets 4 and 5. In treatments with 65% water, the picture was less clear, with no statistical difference in viscosity or elastic modulus values between diets 3, 4 and 5. Diet 1 and diet 2 differed in viscosity but not in elastic modulus. In some cases formulations with combined binder treatment more than compensated for the effect of greater dilution, for example the viscosities of diets 3, 4 and 5 with 62.5% water exceeded those of the control diet with 60% water (Fig. 3).

Fig. 3.

Overview of dynamic viscosity values (means with standard deviation) at measured moisture levels.

Fig. 4.

Dynamic viscosity (mean values with their standard errors) of 5 different diet treatments as a function of angular frequency (model chyme with a water content of 62.5%) and the linear fits of the two power law regions.

3.2. Frequency sweeps

Examples of frequency sweeps for preparations of all five diets in which water content was 62.5% are shown in Fig. 4. The model chyme exhibited consistent strong shear thinning behavior for all treatments and water contents. Visual inspection of the data revealed two power law regions (PLR), in which different treatments yielded differences in the steepness of curves (from measuring point 1 to 8 and 9 to 16). To quantify these differences, the negative slopes of linear fits for each of the two regions were analysed. Smaller negative slope values indicate greater mechanical resistance to the applied stress [5]. The statistical analyses of PLR slopes and their sum values (as measures of shear resistance) are shown in Table 3. In all treatments the mean slope values of PLR 1 exceed those of the corresponding PLR 2. The three diets with the greatest low-shear viscosities (diet 3, 4 and 5) also exhibited the highest shear thinning at all tested moisture levels. In the control treatment, shear thinning was less pronounced at each moisture level than in the corresponding guar gum treatment.

Table 3
Mean values of the slopes of linear fits of the two power law regions (PLR 1, PLR 2) for frequency sweeps and the sum of slopes (mean values with their standard errors) at different water contents without and with inclusion of respective non-starch polysaccharide(s) (NSP)

Diet NSP PLR 1^Ψ PLR 2^Ψ Sum of Slopes^Ψ Water content (wt%)

1 – −0.8869^h −0.6399^e −1.5268^ef ± 0.0069 60%

2 GG −0.8447^ef −0.6931^gh −1.5378^ef ± 0.0043

3 $XG + GG$ −0.8816^h −0.7130^ij −1.5945^h ± 0.0031

4 $XG + TG$ −0.8856^h −0.7184^j −1.6040^h ± 0.0036

5 $XG + LBG$ −0.8772^gh −0.7152^ij −1.5925^gh ± 0.0027

1 – −0.8260^e −0.5838^b −1.4099^bc ± 0.0083 62.5%

2 GG −0.7946^d −0.6643^f −1.4590^d ± 0.0063

3 $XG + GG$ −0.8580^fg −0.6983^hi −1.5564^fg ± 0.0027

4 $XG + TG$ −0.8404^ef −0.6768^fg −1.5172^e ± 0.0032

5 $XG + LBG$ −0.8444^ef −0.6807^fg −1.5252^ef ± 0.0040

1 – −0.7518^b −0.5120^a −1.2638^a ± 0.0553 65%

2 GG −0.6950^a −0.6026^c −1.2976^a ± 0.0074

3 $XG + GG$ −0.7852^cd −0.6388^de −1.4241^bc ± 0.0063

4 $XG + TG$ −0.7699^bc −0.6227^d −1.3926^b ± 0.0030

5 $XG + LBG$ −0.8006^d −0.6397^e −1.4402^cd ± 0.0046

Model effects^†

Diet * * *

Water content * * *

Interaction * * ***

Diet	NSP	PLR 1^Ψ	PLR 2^Ψ	Sum of Slopes^Ψ	Water content (wt%)
1	–	−0.8869^h	−0.6399^e	−1.5268^ef ± 0.0069	60%
2	GG	−0.8447^ef	−0.6931^gh	−1.5378^ef ± 0.0043
3	$XG + GG$	−0.8816^h	−0.7130^ij	−1.5945^h ± 0.0031
4	$XG + TG$	−0.8856^h	−0.7184^j	−1.6040^h ± 0.0036
5	$XG + LBG$	−0.8772^gh	−0.7152^ij	−1.5925^gh ± 0.0027
1	–	−0.8260^e	−0.5838^b	−1.4099^bc ± 0.0083	62.5%
2	GG	−0.7946^d	−0.6643^f	−1.4590^d ± 0.0063
3	$XG + GG$	−0.8580^fg	−0.6983^hi	−1.5564^fg ± 0.0027
4	$XG + TG$	−0.8404^ef	−0.6768^fg	−1.5172^e ± 0.0032
5	$XG + LBG$	−0.8444^ef	−0.6807^fg	−1.5252^ef ± 0.0040
1	–	−0.7518^b	−0.5120^a	−1.2638^a ± 0.0553	65%
2	GG	−0.6950^a	−0.6026^c	−1.2976^a ± 0.0074
3	$XG + GG$	−0.7852^cd	−0.6388^de	−1.4241^bc ± 0.0063
4	$XG + TG$	−0.7699^bc	−0.6227^d	−1.3926^b ± 0.0030
5	$XG + LBG$	−0.8006^d	−0.6397^e	−1.4402^cd ± 0.0046
		Model effects^†
Diet	***	***	***
Water content	***	***	***
Interaction	***	***	***

^Ψ

$Adjusted means \pm standard error (SE)$ . Values within a column not sharing a superscript letter are significantly different.

(Tukey–Kramer HSD; $p < 0.05$ ).

^†

Model effects: *** $p < 0.0001$ .

Thus an inverse correlation was observed between frequency sweep data and values for dynamic viscosity $η^{'}$ and elastic modulus $G^{'}$ . Meanwhile, high mean values for viscosity and elastic modulus indicated lower shear resistance for the measured frequency range. A polynominal regression of slopes versus adjusted means of viscosity (not shown) and elastic modulus (Fig. 5) provides the best fits, at $R^{2} = 0.93$ and $R^{2} = 0.97$ , respectively. For both viscosity and elastic modulus, the relationships between slope and mean values are exponential up to 100 Pa*s and 200 Pa respectively.

Fig. 5.

Polynominal regression (2nd degree) of the adjusted means of elastic modulus as a function of the sum of negative slopes of the two separately fitted regions (resulting from linear fits of the frequency sweeps).

Fig. 6.

Creep test curves of model chyme based on Diet 2 (0.2% GG) and Diet 3 (0.67% GG, 1.33% XG) at three measured water contents (mean values with their standard errors) showing power law relations of the basic viscoelastic models described by Maxwell and Kelvin-Voigt.

3.3. Creep test

Figure 6 shows representative creep curves of two different diet treatments. In general, deformation increases strongly with water content and clear differences between binder treatments are apparent. Plotted on double logarithmic scales for all diet treatments and water contents, the data exhibit a power-law relationship in a first approximation, with a range of exponents between about 1.2 and 1.4. Closer examination of the data indicates two regions worthy of further investigation elsewhere. All curves were plotted on double logarithmic scales (plots not shown) and all fits (according to formula (1)) showed sufficient accuracy ( $R^{2} > 0.99$ ). The two fit parameters A and a were affected by both factors diet and water content and by the interaction between factors ( $p < 0.001$ ), but post hoc analyses of the fit parameters did not show a clear picture (data not shown). A strong inverse correlation was observed between fit parameter A and viscosity values (Fig. 7). High values of fit parameter A indicate a sharp increase in deformation in the creep curves, corresponding with a low mean viscosity in the chyme sample being tested.

Fig. 7.

Fit parameter A as a function of viscosity (adjusted mean values) of the time sweeps.

4. Discussion

The material heterogeneity of stomach contents from real fish stems from individual variation and from differences in the state of digestion, and this variability makes analysis problematic. In vivo, salmonid chyme consists of a relatively homogeneous paste fraction mixed with individually variable particles (not shown). The difficulty in controlling this particle fraction of digesta renders it impossible to reliably characterize untreated chyme by means of oscillatory rheometry. Thus the current attempt to characterize the properties of gastric material in cultured fish used an artificial model of chyme, developed from dry fish feeds. The in vitro models were pre-homogenized and formulated at three different dilutions within a physiologically realistic range.

Despite the relatively complex pretreatment processes of feed manufacture and model chyme production, the rheological measurements showed fairly high reproducibility for all treatments in this study. In case of the time sweeps, the adjusted mean values of viscosity and elastic modulus proved to be suitable parameters for detecting and quantifying even minor changes in feed composition.

4.1. Fundamental and mechanical characterization of in vitro chyme

Viscosity responses showed a time dependent decline and approached a plateau value after 150 s, with only slight further reductions recorded after this point. This decline over time is in accordance with previous studies on the rheology of faeces of rainbow trout [5,22,23]. However, the present results diverged from those previous studies with regard to elastic modulus values, showing a consistent increase over time. The measurements in this study were conducted under conditions yielding data within the LVE-region, applying a small deformation of 0.1%, whereas Brinker et. al [5] applied substantially larger deformations of 30%, based on shear scenarios previously observed in real farming environments. These large deformations illustrated differences in the elastic modulus but did not reveal basic differences in the viscosities of the material. Using small deformations in the LVE-region of the current study ensured that structural disturbances to the material were minimal, and allowed the chyme to show structural recovery after the unavoidable stress caused by transferring samples to the rheometer.

All measured formulations exhibited a pseudo-plastic behavior, with decreasing viscosity as frequency increased. Such behavior is not unusual for polymer solutions [24], and similar strong shear thinning behavior has previously been observed for feed mixtures and fish faeces [5,25,26].

Brinker et al. [5] recorded clear differences in the mean viscosities of faeces from rainbow trout fed diets with different protein sources, but the frequency sweeps conducted as part of that study revealed no differences in shear resistance between investigated protein treatments. It is likely that the large deformations of 30% applied during the preliminary time sweeps in that study resulted in structural alterations to the original faecal material that masked potential differences in shear resistance. The very small deformations used in the present study remained within the LVE-region and did thus not affect the basic structure of the material. The results across the frequency range applied in this study indicate that differences in shear resistance are more pronounced for the more stable materials with mean viscosities greater than 100 Pa*s and elastic modulus values of 200 Pa. Some potential consequences of these differences are described below.

The power law exponents of the creep curves between 1.2 and 1.4 ranged between those described by two basic viscoelastic models, with deformations somewhere between those expected for a viscoelastic fluid as described by the Maxwell model and a viscoelastic solid as described by Kelvin-Voigt [27]. However, the creep response indicates a material distinctly closer to a viscoelastic solid (Fig. 6). Although equation (1) was well able to describe all data fits for the creep curves with a certainty measure in excess of 99% (for restrictions, see below), the fine structure of the creep response is more complex than appears at first sight. For the samples with the highest water contents, deformation increased strongly after 0.5 Pa, indicating a regime change towards fluid-like behavior due to structural alterations of the material. For this reason, plots of all samples containing 65% water were limited to 0.5 Pa to ensure comparability. Brinker et al. [5] pursued creep experiments with faeces of rainbow trout until a complete material breakdown occurred, but despite the far greater range of stress applied, the material behavior showed similar progress, with a linear phase followed by a strong increase in deformation before breaking down. Figure 7 shows a strong correlation between fit parameter A and the mean values of dynamic viscosity $η^{'}$ in time sweeps with an $R^{2}$ of 0.96. This is a consequence of the linear viscoelastic properties under which the experiments were conducted and applies equally to the earlier frequency sweeps, as confirmed by the good correlation of the linear fit slopes with the observed means of $η^{'}$ . The highly predictable fit parameter A renders the creep experiment theoretically redundant for stability analyses, but the observed upturn in the high moisture samples after 0.5 Pa indicates a material response outside the LVE-region which should be investigated further if material properties of chyme during digestion of commercial extruded feeds by farmed salmonids are to be fully understood.

4.2. Implications of material properties for digestive physiology and potential consequences for aquaculture operations

Extruded pellets are not comparable to natural fish diets and artificial feeds inevitably present some physiological challenges to the fish consuming them. Artificial feeds typically contain 8–10% water compared to the 70–80% in prey ingested by wild fish [28]. Thus the stomach content of fish consuming extruded diets requires considerable active or passive liquefaction by the fish. For rainbow trout there seems to be a physiological limit to the amount of moisture that can be added within the stomach, amounting to a final water content of about 65% [15,17,18,29].

As discussed, the consistency of digesta and faeces depends on feed composition and in particular on the proportions of rheologically active ingredients. Vegetable protein sources contain variable quantities of NSPs depending on source species and strain, processing and environmental factors and post-harvest storage conditions [30], and some NSPs are known to confer high viscosity (e.g. β-glucans). In addition to naturally occurring thickeners, binders are increasingly used in commercial aqua feeds in order to manipulate the properties of faeces. For example, the inclusion in feed of 0.3% by weight of guar gum significantly increases faecal stability [31]. The present study is therefore pertinent on several practical fronts.

As gastric and intestinal motility seem to be more or less insensitive to changes in the viscosity of digesta [32], the efficiency of mixing in the GI tract is closely related to the composition of ingested feed. It has been shown that increased viscosity of gastric chyme can impact homogenization and evacuation time in monogastric animals, including fish [33], and hamper enzyme substrate interactions [34]. Numerous studies indicate a reduction in nutrient digestibility in diets containing GG or other NSPs, for example, lower protein digestibility observed in broiler chickens and rats and fish including rainbow trout, salmon, tilapia and African catfish was wholly or partly attributed to higher viscosity chyme [33,35–37]. The model chyme developed for this study was based on physiological conditions in the stomach but several studies have shown that viscosity in the anterior intestines is strongly related to gastric viscosity and as most rheologically active components are enzymatically indigestible for salmonids, they accumulate during digestion and their viscosity enhancing effect is intensified [38,39]. It cannot be assumed that elevated viscosity of digesta necessarily impacts digestion in a negative way. Some highly effective binders have been trialed without negative effects on digestion or feed performance [23,31,40], but might in combination result in chyme consistencies that exceed the physiological capabilities of the fish [37,41].

The differences in viscosity and elastic modulus between treatments were particularly pronounced in the samples containing only 60% water, and far exceeded expectations. While the quantity of binder incorporated into commercial aqua feeds is calculated to arrive at an effective concentration (e.g. 0.2% wt for guar gum) after dewatering and compaction in the faecal cast and thus takes account of the accumulation of indigestible polysaccharides during digestion [5], the rheological effects are likely to be noticeable much earlier, during the gastric phase of digestion. Even so, the observed fourfold increases in viscosity and elastic modulus values for binder treated model chyme over control material in the present study were surprising and cannot be explained by the thickening properties of cold water. The inconsistent hierarchy of binder effects between samples with different moisture levels must be attributed to a combination of factors, including the specific hydrating properties of the binder compounds and the influence of concentration on synergistic binder effects.

The thickening properties of polysaccharide hydrocolloids deployed singly and sparingly are mainly based on molecular weight, molecular structure (e.g. degree of branching) and concentration [42]. In addition, the NSP-combinations in this study were selected on the basis of the well-known synergistic interactions between xanthan and polysaccharides of the galactomannan family, in order to be effective at low concentration. The intensity of this synergism is thought to correlate positively with the number of “smooth regions” on the mannose backbone, which lacks the galactose stubs of the respective galactomannan species [43]. This may be one explanation for the significantly higher viscosity values of $XG + LBG$ compared to $XG + GG$ in the 60% water samples. The mannose:galactose ratio of LBG is 1:4, compared to 1:2 for guar gum [44] and the interaction between the two binders is thought to be based on the association of xanthan side chains with the galactomannan backbones [14]. $XG + TG$ showed highest values for both viscosity and elastic modulus, despite reports in the literature that their synergism is less pronounced than that of XG and LBG [13], (the mannose:galactose ratio of tara gum is 1:3). Thus far therefore, an adequate synergistic explanation for the phenomenon between these polysaccharides is lacking.

While the dilution of treatments implies a reduction in the influence of synergism, other specific properties of single NSPs, such as molecular weight and capacity to bind water, remain highly pertinent to viscosity. The water binding capacity of xanthan gum is about 6 times greater than that of guar gum and 20 times that of locust bean gum [45], which may at least partly explain the relatively high viscosity values of XG mixes. Xanthan gum has a relatively low molecular weight and a low cold water viscosity compared to guar gum, and interactions in the observed matrix become more complex as moisture levels are reduced. It is also important to consider that these individual binders or binder combinations are additives to what is already a highly complex and variable matrix, whose rheological character is increasingly unpredictable. Sánchez et al. [45] found synergistic effects of both xanthan and guar gum with soy protein, a crucial ingredient in aqua feed. For soy protein-xanthan mixtures, matrix formation is attributed to the strongly hydrophilic nature of xanthan gum, leading to a high water holding capacity of the soluble complex and an increase in internal cohesion [46]. The $XG + GG$ treatment formulated with 62.5% water resulted in the highest observed viscosity at the same water level, lending possible support to the synergistic theory described above.

Further complex interactions are also known between NSP gums and starch from various sources [47]. Considering of the diversity of potential interactions in the feed matrix, it is difficult to condense outcomes into simple rules. Binder synergism is a mixed blessing. On one hand, carefully selected NSPs might be used to enhance faecal stability well over what can be achieved by single binders of the same concentration, while on the other hand the potential for negative side effects of enhanced gastric viscosity on digestion is large. Obvious potential problems include increased micelle size [48] or impaired diffusion [39], both of which might reduce exposure of nutrients to digestive enzymes [49] and thus severely hamper digestion. The sensitivity of the rheological approach taken in this study offers an opportunity to improve understanding of how the consistency of chyme affects digestive processes including nutrient digestion and uptake and to investigate, review and ultimately predict the digestibility effects of novel plant feed ingredients or combinations.

Previous studies of physiological shear rates in the GI tract are limited. A value of 0.03 to 10 s⁻¹ was reported for small mammals [50], and contraction frequencies of 0.01 to 12 min⁻¹ are reported for fish gut [51–53]. These data are as yet insufficient to formulate a useful classification of shear resistance, but there is some evidence that increases in low-shear viscosities may inhibit digestive mixing. In a study of in vitro digestion of starch, Hardacre et al. [54] describe a threshold shear rate necessary to achieve adequate mixing of chyme. In the present study, the three diets with the highest low-shear viscosities (diet 3, 4 and 5) also exhibited the strongest shear-thinning. Considering the potential for inhibition of digestive mixing, this material feature may help to explain the apparent lack of consequences in the GI tract resulting from the inclusion of rheologically active feed additives. Further investigations are needed to identify physiological shear conditions and forces within the GI tract of fishes and to assess the extent to which modern feeds address this potential physiological challenge to the welfare and performance of farmed fish.

A little more is known about physiological shear forces occurring in the GI tracts of monogastric animals. The maximum shear stress of 1 Pa applied in the creep tests of the present study is close to the maximum of 1.2 Pa found in guinea pig ileum [55]. Values of about 1 Pa to 2 kPa have been reported in the intestines of some fish species [53,56]. The creep curves of the samples with the highest moisture levels showed a strong increase in deformation when shear stresses exceed 0.5–0.6 Pa, indicating a loss of material stability. The water contents of natural chyme peaks in the advanced stages of gastric digestion before the chyme is transported to the intestinal tract, and passage through the pylorus is known to be consistency dependent [57]. Delayed passage is thus a likely consequence of highly viscous digesta, and has the potential to increase satiety, reduce daily feed intake and thus limit growth performance. However, the results of the creep experiments and time sweeps in this study indicate that the final dilution of gastric digesta might be sufficient to prevent critical concentrations (c*) of rheologically active polysaccharides exceeding physiological relevant thresholds.

In the proximal intestine, the water content of chyme may rise to over 80% at the point where nutrient resorption and accumulation of indigestible substances sets in [15]. The role of viscosity-enhancing materials may be subordinate in this part of the GI tract but becomes more significant with the increasing concentration of rheologically active substances that follows the uptake of the digestible fraction.

The established model represents a novel tool for investigating the rheological consequences of changing feed compositions on both gastrointestinal and faecal chyme. This methodology opens a new avenue of chyme research, with implications for fish nutrition, physiology, welfare and performance. Regardless of the complexity of rheological interactions in the feed matrix, digesta, chyme and faeces from all areas of the GI tract are affected, since the relevant substances are largely indigestible and therefore accumulate during digestion. This concentration effect is likely to enhance both intentional and unintended effects on chyme consistency as digestion proceeds.

A valuable further step would be to incorporate measured parameters of the model chyme and faeces into a faecal stability prediction model based solely on feed composition. This would be a highly valuable tool with which to tailor feed composition to optimize water quality and fish health.

Furthermore, the simplified model can serve as a means of investigating other factors likely to influence the flow properties of chyme, such as pH or enzyme activity, and thus to further improve understanding of the fundamental rheological processes at work in the GI tract.

Funding

This project was funded by Deutsche Bundesstiftung Umwelt (DBU) (AZ 26128). DBU had no role in the design, analysis or writing of this article.

Author contributions

The contributions of the authors were as follows: The study was designed by M.S., A.B., C.F. Data collection and drafting the article was done by M.S. All authors contributed to the analysis and interpretation of the data and approved the final version of the manuscript to be published. The authors declare no competing financial interest.

Footnotes

Acknowledgements

The authors are very grateful to Ulrich Matthes for the excellent technical assistance and lab work. We are very grateful to J. Holm (Biomar) for providing the basal mix of the experimental diets. We thank Carina Osterwinter for her help with the preparation and the set-up of the measuring system. A special thank goes to Katharina Steinbrenner for the graphical support. The paper benefited greatly from editing of the scientific language by A.-J. Beer.

References

FAO. The State of World Fisheries and Aquaculture – 2012 (SOFIA). Rome; 2012.

Ytrestøyl

Aas

Åsgård

. Utilisation of feed resources in production of Atlantic salmon (Salmo salar) in Norway. Aquaculture. 2015;448:365–374. doi:10.1016/j.aquaculture.2015.06.023.

Krogdahl

Gajardo

Kortner

Penn

Berge

, et al. Soya saponins induce enteritis in Atlantic salmon (Salmo salar L.). J Agric Food Chem. 2015;63:3887–3902. doi:10.1021/jf506242t.

Krogdahl

Penn

Thorsen

Refstie

Bakke

. Important antinutrients in plant feedstuffs for aquaculture: An update on recent findings regarding responses in salmonids. Aquac Res. 2010;41:333–344. doi:10.1111/j.1365-2109.2009.02426.x.

Brinker

Friedrich

. Fish meal replacement by plant protein substitution and guar gum addition in trout feed. Part II: Effects on faeces stability and rheology. Biorheology. 2012;49:27–48.

Davidson

Good

Barrows

Welsh

Kenney

Summerfelt

. Comparing the effects of feeding a grain- or a fish meal-based diet on water quality, waste production, and rainbow trout Oncorhynchus mykiss performance within low exchange water recirculating aquaculture systems. Aquac Eng. 2013;52:45–57. doi:10.1016/j.aquaeng.2012.08.001.

Unger

Brinker

. Feed and treat: What to expect from commercial diets. Workshop Recirc Aquac Syst. 2013;53:19–29.

Badiola

Mendiola

Bostock

. Recirculating aquaculture systems (RAS) analysis: Main issues on management and future challenges. Aquac Eng. 2012;51:26–35. doi:10.1016/j.aquaeng.2012.07.004.

Becke

Steinhagen

Schumann

Brinker

. Physiological consequences for rainbow trout (Oncorhynchus mykiss) of short-term exposure to increased suspended solid load. Aquac Eng. 2016; In press.

10.

Dalsgaard

Pedersen

. Solid and suspended/dissolved waste (N, P, O) from rainbow trout (Oncorynchus mykiss). Aquaculture. 2011;313:92–99. doi:10.1016/j.aquaculture.2011.01.037.

11.

Casas

Mohedano

Garcia-Ochoa

. Viscosity of guar gum and xanthan/guar gum mixture solutions. J Sci Food Agric. 2000;80:1722–1727. doi:10.1002/1097-0010(20000915)80:12<1722::AID-JSFA708>3.0.CO;2-X.

12.

Casas

García-Ochoa

. Viscosity of solutions of xanthan/locust bean gum mixtures. J Sci Food Agric. 1999;79:25–31. doi:10.1002/(SICI)1097-0010(199901)79:1<25::AID-JSFA164>3.0.CO;2-D.

13.

Tako

. Synergistic interaction between xanthan and tara-bean gum. Carbohydr Polym. 1991;16:239–252. doi:10.1016/0144-8617(91)90111-O.

14.

Tako

Nakamura

. Synergistic interaction between xanthan and guar gum. Carbohydr Res. 1985;138:207–213. doi:10.1016/0008-6215(85)85104-1.

15.

Bucking

Wood

. Water dynamics in the digestive tract of the freshwater rainbow trout during the processing of a single meal. J Exp Biol. 2006;209:1883–1893. doi:10.1242/jeb.02205.

16.

Hilton

Cho

Slinger

. Effect of extrusion processing and steam pelleting diets on pellet durability, pellet water absorption, and the physiological response of rainbow trout (Salmo gairdneri R.). Aquaculture. 1981;25:185–194. doi:10.1016/0044-8486(81)90180-0.

17.

Kristiansen

Cliff Rankin

. Discrimination between endogenous and exogenous water sources in juvenile rainbow trout fed extruded dry feed. Aquat Living Resour. 2001;14:359–366. doi:10.1016/S0990-7440(01)01131-7.

18.

Ruohonen

Grove

McIlroy

. The amount of food ingested in a single meal by rainbow trout offered chopped herring, dry and wet diets. J Fish Biol. 1997;51:93–105.

19.

Tunick

. Small-strain dynamic rheology of food protein networks. J Agric Food Chem. 2011;59:1481–1486. doi:10.1021/jf1016237.

20.

Sokal

Rohlf

. Biometry: The principles and practices of statistics in biological research. 3rd ed. New York: W.H. Freeman; 1994.

21.

Hayter

. A proof of the conjecture that the Tukey–Kramer multiple comparisons procedure is conservative. Ann Stat. 1984;12:61–75. doi:10.2307/2241034.

22.

Brinker

. Improving the mechanical characteristics of faecal waste in rainbow trout: The influence of fish size and treatment with a non-starch polysaccharide (guar gum). Aquac Nutr. 2009;15:229–240. doi:10.1111/j.1365-2095.2008.00587.x.

23.

Brinker

Koppe

Rösch

. Optimised effluent treatment by stabilised trout faeces. Aquaculture. 2005;249:125–144. doi:10.1016/j.aquaculture.2004.12.029.

24.

Picout

Ross-Murphy

. Rheology of biopolymer solutions and gels. ScientificWorldJournal. 2003;3:105–121. doi:10.1100/tsw.2003.15.

25.

Kannadhason

Muthukumarappan

Rosentrater

. Effects of ingredients and extrusion parameters on aquafeeds containing DDGS and tapioca starch. J Aquac Feed Sci Nutr. 2009;1:6–21.

26.

Lam

Flores

. Effect of particle size and moisture content on viscosity of fish feed. Cereal Chem J. 2003;80:20–24. doi:10.1094/CCHEM.2003.80.1.20.

27.

Nijenhuis

McKinley

Spiegelberg

Barnes

Aksel

Heymann

, et al. Non-Newtonian flows. In: Tropea

Yarin

Foss

, editors. Springer handb. exp. fluid mech. Berlin Heidelberg: Springer; 2007. p. 619–743. doi:10.1007/978-3-540-30299-5_9.

28.

Guillaume

Kaushik

Bergot

Metailler

. Nutrition and feeding of fish and crustaceans. 1st ed. Springer; 2001.

29.

Hilton

Atkinson

Slinger

. Effect of increased dietary fiber on the growth of rainbow trout (Salmo gairdneri). Can J Fish Aquat Sci. 1983;40:81–85. doi:10.1139/f83-012.

30.

Kim

Mullan

Simmins

Pluske

. Variation in the chemical composition of wheats grown in western Australia as influenced by variety, growing region, season, and post-harvest storage. Aust J Agric Res. 2003;54:541–550. doi:10.1071/AR02183.

31.

Brinker

. Guar gum in rainbow trout (Oncorhynchus mykiss) feed: The influence of quality and dose on stabilisation of faecal solids. Aquaculture. 2007;267:315–327. doi:10.1016/j.aquaculture.2007.02.037.

32.

Lentle

Janssen

PWM

. Physical characteristics of digesta and their influence on flow and mixing in the mammalian intestine: A review. J Comp Physiol [B]. 2008;178:673–690. doi:10.1007/s00360-008-0264-x.

33.

Choct

Hughes

Wang

Bedford

Morgan

Annison

. Increased small intestinal fermentation is partly responsible for the anti-nutritive activity of non-starch polysaccharides in chickens. Br Poult Sci. 1996;37:609–621. doi:10.1080/00071669608417891.

34.

Hardacre

Yap

S-Y

Lentle

Monro

. The effect of fibre and gelatinised starch type on amylolysis and apparent viscosity during in vitro digestion at a physiological shear rate. Carbohydr Polym. 2015;123:80–88. doi:10.1016/j.carbpol.2015.01.013.

35.

Leenhouwers

Adjei-Boateng

Verreth

Schrama

. Digesta viscosity, nutrient digestibility and organ weights in African catfish (Clarias gariepinus) fed diets supplemented with different levels of a soluble non-starch polysaccharide. Aquac Nutr. 2006;12:111–116. doi:10.1111/j.1365-2095.2006.00389.x.

36.

Refstie

Svihus

Shearer

Storebakken

. Nutrient digestibility in Atlantic salmon and broiler chickens related to viscosity and non-starch polysaccharide content in different soyabean products. Anim Feed Sci Technol. 1999;79:331–345. doi:10.1016/S0377-8401(99)00026-7.

37.

Storebakken

. Binders in fish feeds: I. Effect of alginate and guar gum on growth, digestibility, feed intake and passage through the gastrointestinal tract of rainbow trout. Aquaculture. 1985;47:11–26. doi:10.1016/0044-8486(85)90004-3.

38.

Gatlin

Barrows

Brown

Dabrowski

Gaylord

Hardy

, et al. Expanding the utilization of sustainable plant products in aquafeeds: A review. Aquac Res. 2007;38:551–579. doi:10.1111/j.1365-2109.2007.01704.x.

39.

Sinha

Kumar

Makkar

HPS

De Boeck

Becker

. Non-starch polysaccharides and their role in fish nutrition – A review. Food Chem. 2011;127:1409–1426. doi:10.1016/j.foodchem.2011.02.042.

40.

Brinker

Reiter

. Fish meal replacement by plant protein substitution and guar gum addition in trout feed, Part I: Effects on feed utilization and fish quality. Aquaculture. 2011;310:350–360. doi:10.1016/j.aquaculture.2010.09.041.

41.

Storebakken

Austreng

. Binders in fish feeds: II. Effect of different alginates on the digestibility of macronutrients in rainbow trout. Aquaculture. 1987;60:121–131. doi:10.1016/0044-8486(87)90304-8.

42.

Williams

Phillips

Sworn

. Hydrocolloid thickeners and their applications. Gums Stabilisers Food Ind. 2004;12:13–22. doi:10.1039/9781847551214-00013.

43.

Morris

. Gelation of polysaccharides. In: Functional properties of food macromolecules. Gaithersburg, Maryland: Aspen Publishers; 1998. p. 202–203.

44.

Benichou

Aserin

Garti

. Protein-polysaccharide interactions for stabilization of food emulsions. J Dispers Sci Technol. 2002;23:93–123. doi:10.1080/01932690208984192.

45.

Sánchez

Bartholomai

Pilosof

AMR

. Rheological properties of food gums as related to their water binding capacity and to soy protein interaction. LWT – Food Sci Technol. 1995;28:380–385.

46.

Hua

Cui

Wang

. Gelling property of soy protein–gum mixtures. Food Hydrocoll. 2003;17:889–894. doi:10.1016/S0268-005X(03)00110-3.

47.

Heyman

De Vos

Van der Meeren

Dewettinck

. Gums tuning the rheological properties of modified maize starch pastes: Differences between guar and xanthan. Food Hydrocoll. 2014;39:85–94. doi:10.1016/j.foodhyd.2013.12.024.

48.

Meyer

Doty

. GI transit and absorption of solid food: Multiple effects of guar. Am J Clin Nutr. 1988;48:267–273. doi:10.1093/ajcn/48.2.267.

49.

Smulikowska

Mieczkowska

Nguyen

. The influence of digesta viscosity on the development of the stomach, on in vitro small intestinal motility and on digestion of nutrients in broiler chickens. J Anim Feed Sci Pol. 2002.

50.

de Loubens

Lentle

Love

Hulls

Janssen

PWM

. Fluid mechanical consequences of pendular activity, segmentation and pyloric outflow in the proximal duodenum of the rat and the Guinea pig. J R Soc Interface. 2013;10:20130027. doi:10.1098/rsif.2013.0027.

51.

Brijs

Hennig

Axelsson

Olsson

. Effects of feeding on in vivo motility patterns in the proximal intestine of shorthorn sculpin (Myoxocephalus scorpius). J Exp Biol. 2014;217:3015–3027. doi:10.1242/jeb.101741.

52.

Gräns

Albertsson

Axelsson

Olsson

. Postprandial changes in enteric electrical activity and gut blood flow in rainbow trout (Oncorhynchus mykiss) acclimated to different temperatures. J Exp Biol. 2009;212:2550–2557. doi:10.1242/jeb.030593.

53.

Olsson

Aldman

Larsson

Holmgren

. Cholecystokinin affects gastric emptying and stomach motility in the rainbow trout Oncorhynchus mykiss . J Exp Biol. 1999;202:161–170.

54.

Hardacre

Lentle

Yap

S-Y

Monro

. Does viscosity or structure govern the rate at which starch granules are digested? Carbohydr Polym. 2016;136:667–675. doi:10.1016/j.carbpol.2015.08.060.

55.

Jeffrey

Udaykumar

Schulze

. Flow fields generated by peristaltic reflex in isolated Guinea pig ileum: Impact of contraction depth and shoulders. Am J Physiol – Gastrointest Liver Physiol. 2003;285:G907–G918. doi:10.1152/ajpgi.00062.2003.

56.

Karila

Holmgren

. Enteric reflexes and nitric oxide in the fish intestine. J Exp Biol. 1995;198:2405–2412.

57.

Grosell

Farrell

Brauner

. Fish physiology: The multifunctional gut of fish. vol. 30. Academic Press; 2010.