Physicochemical properties of dietary protein as predictors for digestibility or releasing percentage of amino acids in monogastrics under in-vitro conditions

Abstract The nutritional values of the protein are related to its amino acid content, and its physicochemical properties. We aimed to determine the relationship between the variation of protein digestibility or the releasing coefficient of amino acids with the physicochemical properties of different feedstuffs. We analysed the secondary protein structure and solubility of 10 protein-rich feed ingredients, including soybean, soybean meal, dehulled soybean meal, fermented soybean meal, distiller’s dried grains with solubles (DDGS), feather meal, fish meal, rapeseed meal, corn gluten meal, and corn germ meal. The results showed that the soybeans and their products had greater protein solubility than the other dietary protein (p < .001). Fish meal, rapeseed meal and all soybean sources had the higher protein digestibility and releasing coefficient of amino acids, also reached different peaks in shorter times. But corn germ meal, corn gluten meal, and feather meal exhibited lower releasing coefficient of amino acids and weaker solubility (p < .001). The in-vitro protein digestibility or releasing coefficient of total amino acid had positive regression coefficients with α-helix, α-helix/β-sheet, and solubility; and a negative regression coefficient with β-sheet. Feedstuffs with faster amino acid releasing rate had higher α-helix, α-helix/β-sheet and solubility and lower β-sheet. Therefore, the physicochemical properties of the dietary protein could be used as predictors for the amino acids’ digestibility and amino acid bioavailability. Highlights We evaluated the protein solubility, protein digestibility, and releasing coefficient of amino acids among the most common feedstuffs for monogastric animals The physicochemical properties of dietary protein could be used as predictors for the digestibility or releasing rate of amino acids in monogastric. The feed industry formulation might be reconsidered since the most common feedstuffs possess novel concepts.


Introduction
Soybeans and their products are widely used throughout the world as a dietary protein source due to their high protein content and well-balanced amino acid (AA) (Friedman and Brandon 2001). Nowadays, many countries import soybean meal to feed different animal species, which increases the economic cost of animal feeding (Karlsson et al. 2021). Many studies have been conducted to investigate alternative protein sources as a partial replacement, instead of using soybean. These protein sources include oilseed meals, distiller's dried grains with solubles (DDGS), corn gluten meal, and animal protein sources. Some studies have shown that replacing soybean meal with rapeseed meal or faba bean had no adverse effects on the growth performance of growing-finishing monogastric such as pigs (Partanen et al. 2003;Skoufos et al. 2016). However, the excess protein consumption than the normal requirements or the low protein digestibility increases the excretion of nitrogen, which consequently may negatively affect the environment Li et al. 2018). Moreover, the imbalance of protein components leads to the unstable release of AA during digestion and poor utilisation (Reidy et al. 2013;Reidy et al. 2014), which also represents excessive waste. Therefore, it is important to formulate dietary protein that meets protein and AA requirements for the monogastric and to minimise the costs of diets and the excretion of excessive nitrogen into the environment at the same time.
Although there are many kinds of protein sources, not all of them are of equal quality. Processing of soybean meal such as fermentation and puffing could inactivate anti-nutritional factors and even improve protein utilisation (Woodworth et al. 2001;Bao et al. 2014;Jeong et al. 2016). The miscellaneous meals with their high fibre and antinutritional factors, such as cottonseed meal and rapeseed meal can be supplemented in limited ratios (Pelitire et al. 2014;Torres-Pitarch et al. 2014). Because of the poor utilisation of DDGS in pigs, the DDGS can be supplemented to approximately 10-20% ). The body weight was decreased with increasing the levels of DDGS from 0 to 20% (Avelar et al. 2010) or from 0 to 30% (Agyekum et al. 2014) in growing monogastric feed, despite this feedstuff possesses similar net energy and standard ileal digestibility amino acids in the monogastric (Liao et al. 2015;Matthews 2020). The feather meal is limited in lysine content, so it has not been extensively used for humans or other monogastric animals (Chiba et al. 1996;Matthews 2020). Therefore, the nutritional properties of protein feedstuffs should be investigated to optimise poor protein sources and to make up for the scarcity of superior protein sources.
The previous studies have indicated that different protein sources with the same levels of protein and amino acid possess a different effect on the growth performance and nitrogen utilisation for monogastric animals (Zhong et al. 2017). Although the feeds are formulated to contain similar net energy and standard amino acids digestibility, the kinetics of digestion and the release of the amino acids still depend on the dietary protein source (Ma et al. 2019). The recent findings showed that the releasing percentage of amino acids from dietary protein can directly affect the rate of protein synthesis in muscles (Kanda et al. 2016;Berrazaga et al. 2020). This relation reveals the dynamic releasing of amino acids and small peptides in the digestive tract, which is related to the deposition of nitrogen. Therefore, the potential cause of the difference in the growth performance and nitrogen utilisation for monogastric animals might be attributed to the various release of amino acids during digestion and absorption of protein diets, which leads to the imbalance and asynchrony of amino acid entering the amino acid metabolic pool in vivo. It is vital to make a clear temporal and spatial variation of amino acid in the gastrointestinal tract for the various protein feedstuffs in order to precisely formulate the diets.
The protein digestibility can be affected by different factors related to the structural characteristics of each protein. (Yu 2010;Samadi et al. 2013;Peng et al. 2014;Yan et al. 2014;Bai et al. 2015). Such factors include the ratio of a-helices and b-sheets (Dyson and Wright 1993), amide I and II bands (Kinsella and Melachouris 1976;Bai et al. 2015); protein solubility (Kinsella and Melachouris 1976), and the particle size of feedstuffs (Xing et al. 2017). Therefore, it is necessary to clarify the relationship between the physicochemical properties of dietary protein ingredients and they are in vitro protein digestibility in order to predict the in vivo digestion and absorption of protein. For instance, in vitro experiments are widely performed and are mostly related to the in vivo estimates (Ngalavu et al. 2020). The in vitro digestion method is commonly performed to study structural changes, digestion ability, and liberation of feed components to simulate the gastrointestinal factors by using pigs as a model for the monogastric (Ngalavu et al. 2020). Limited studies have been conducted to compare the digestibility and the releasing rate of amino acids in different proteinrich feedstuffs (Ngalavu et al. 2020). We hypothesised that the physicochemical structure of a feedstuff may be related to its ability to be hydrolysed by digestive enzymes in the monogastric, and consequently can affect protein digestibility and the releasing rate of the amino acids.
This study aims to determine the secondary structure and solubility (%) of different dietary protein sources for monogastric animals, to evaluate the variation in protein digestibility among these dietary protein sources, and to predict the in vitro digestibility and the releasing rate of amino acids based on the physicochemical properties of each dietary protein source. These results could provide important information to maximise the utilisation of dietary protein in the feed formulation.

Materials and methods
This study was conducted according to the Ethical Guideline of Jilin Agricultural University No. KT2019008.

Preparation of feed ingredient samples
Ten protein-rich feed ingredients for pig diets, including soybean, soybean meal, dehulled soybean meal, fermented soybean meal, DDGS, feather meal, fish meal, rapeseed meal, corn gluten meal, and corn germ meal, were provided by Changchun Borui Animal Husbandry Co., Ltd. (Changchun, China; Table 1). The feed ingredients were controlled within less than 5% variation in order to keep their uniformity. All used feed ingredient samples in this experiment were collected from raw commodities and were stored in a well-ventilated, cool (temperature 20 C) and dry place (relative humidity 60%) without sunshine for one month to prevent protein deterioration. The free moisture in soybean was 80.64 ± 6.53; in soybean meal, 85.32 ± 7.04; in dehulled soybean meal, 85.19 ± 6.91; in fermented soybean meal,80.17 ± 7.22;in DDGS,100.28 ± 9.24;in feather meal,84.89 ± 5.84;in fish meal,90.13 ± 7.43;in rapeseed meal,90.32 ± 8.12;in corn gluten meal,75.24 ± 5.71; and in corn germ meal was 74.78 ± 6.14 g/kg as an air-dry basis. Each feed ingredient sample was crushed in high-speed multifunction crushing (FW100, Tester, Tianjin, China), passed through a 1 mm sieve, and was stored at 4 C for subsequent analysis. Each chemical analysis was conducted with 3 replicates per feed ingredient. Dry matter (AOAC 1996;method 930.15), crude protein (AOAC 1996;method 984.13), and crude fibre (AOAC 1996;method 978.10), ether extract (AOAC 1996;method 920.39), and ash (AOAC 1996; method 942.05) of feedstuffs were determined using standard methods. Free amino acids were analysed by Ultra highperformance liquid system following hydrolysis seen from the following method of the present study. Briefly, dry matter was determined by oven drying at 105 C for 24 h; Crude protein was analysed by a fully automatic Kjeldahl nitrogen 2300 analyser unit (Foss, Hilleroed, Denmark); Crude fibre was analysed by a 200 Fibre Analysers (ANKOM, NY, USA). Amino acids of all samples were analysed following acid hydrolysis using an Ultra high-performance liquid system (Waters, Milford, MA, USA). Especially, performic acid oxidation was performed prior to hydrolysis to oxidise cystine and methionine. The detailed condition of hydrolysis was conducted according to Chen et al. (2016). DDGS: dried grains with solubles.

Determination of protein solubility
Solubility was determined according to the method of Dale et al. (1987). Briefly, a sample of 1.5 g of each 10 dietary protein ingredients and a 75 mL of 0.2% (w/v) potassium hydroxide solution were mixed in a 100 mL conical flask. Then the mixture was stirred for 20 min. Finally, the supernatant was collected after being centrifuged at 2700 xg for 10 min. Determination of protein content in the supernatant was performed by the Kjeldahl method (by multiplying the N value with 6.25) (AOAC 1996(AOAC , 2002. Protein solubility (PS, %) was calculated as the following PS % ¼ CP concentrations in the supernatant CP content in the original sample Â 100:

Determination of protein secondary structure
Fourier transform infra-red (FTIR) analysis of ten protein feed ingredients was performed according to the method as described by Bai et al. (2015). Briefly, 2 mg sample and 200 mg KBr were mixed and dried in a vacuum dryer for 24 h. The mixture was uniformly ground and pressed into a translucent tablet with 0.25 mm thickness using an infra-red tablet press (FW-4, Thermo Fisher Scientific, Waltham, MA, USA). FTIR spectra (FTIR-8400s, Shimadzu, Kyoto, Japan) were obtained in the mid-infra-red range from 4000 to 400 cm À1 at a spectrum resolution. A total of 64 scans with 4 cm À1 resolution was acquired for each spectrum. A KBr pellet without protein was considered as a control. The infra-red spectra were processed using OMNIC 8.0 software (Nicolet Analytical Instruments, Madison, WI, USA). The baseline of obtained spectra was corrected through control spectra. Origin 7.5 software (Origin Lab, Northampton, MA, USA) was used for Fourier self-deconvolution and Gaussian curve fitting in the region of the amide I band to separate overlapping bands. The secondary structure content of samples was analysed from infra-red second-derivative amide I spectra by manually computing the relative peak areas under the bands that were assigned to a particular substructure.

Determination of the in vitro protein digestibility
In vitro pepsin-trypsin, a two-step digestion experiment was conducted according to Boisen and Fern aNdez (1995). A sample of 1.0 g (as an air-dry basis) from each feed ingredient was added to 10 mL of 0.01 M hydrochloric acid (pH 2.0) and pepsin solution containing 1.0 mg porcine pepsin ( 8, 12, 16, 20, 24, and 28 h. At the end of each collection time, 5 mL of 20% sulfosalicylic acid was added and the samples were centrifuged at 15000 xg for 15 min. Then, 1 mL of the filtration membrane was collected into a tube and centrifuged at 14500 xg for 15 min. The supernatant was collected and centrifuged at 14500 xg for 5 min. The supernatant was then collected and passed through a 0.22-0.45 lm filter, and the filtrate was stored in an À80 C freezer. The precipitate was heated at 80 C for 24 h, and then we used the Kjeldahl method with nitrogen to a protein conversion factor of 6.25 (AOAC 2002), to estimate the crude protein (CP) concentration in the precipitate. According to Wiseman (2018), the in vitro accumulated protein digestibility coefficient (APDC) was calculated using the following formula: The in vitro protein digestibility coefficient per unit time (PDCPUT) was calculated using the following formula: The final APD subtract the initial APD per unit time per unit time

Determination of the in vitro releasing rate of amino acid
We determined the concentration of free amino acid from the obtained supernatant during the in vitro digestion experiment, using the Waters ACQUITY Ultra high-performance liquid TUV system (Waters, Milford, MA, USA) with an AccQÁTagTM Ultra column (2.1 100 mm, P/N: 186003837) using AccQ Tag Ultra Derivatization Kit (Waters, MA, USA). Briefly, 10 lL sample was diluted by10 times (P/NWT007571, Waters, MA, USA) and was treated with 70 lL AccQTag Ultra Borate buffer. Such diluted sample was added to the bottom of a clean derivative bottle. Then, we added 20 lL AccQTag Ultra Derivatization Reagent (1 mL AccQTag Ultra Reagent Diluent is precisely poured into Ultra Derivatization Reagent Powder) in whirling status, and the whirling has lasted for 15 s. After being whirled for 15 s, the mixture was stood for 1 min at room temperature, sealed, and heated at 55 C for a maximum time of 10 min. The derivation product was transferred into a UPLC full recovery sample bottle to be placed in a sample tray. The evaluated parameters included mobile phase A: AccQÁTagUltra eluent A solution, diluted 10 times with pure water, mobile phase B: AccQÁTagUltra eluent B solution, flow rate: 0.7 mL/min; injection volume: 1 lL, column temperature: 55 C, sample temperature: 15 C, detection wavelength: 260 nm, acquisition speed: 20 points /s, time constant: 0.1 s, running time: 10 min.
According to Wiseman (2018), the in vitro accumulated releasing coefficient of an amino acid (ARCAA) was calculated as follows: Ã ARCAA g=g ð Þ ¼

Amino acid content in hydrolysate Amino acid content of the original sample
The in vitro releasing coefficient of amino acid per unit time (RCAAPUT) was calculated as follows: The final ARPAA subtract the initial ARPAA per unit time per unit time

Statistical analysis
All data were analysed using the General Linear Model procedure of the Statistical Package for Social Sciences version 24 (SPSS Inc., Chicago, IL, USA). Each testing data from 3 independent experimental units was used to all statistical analysis. Multiple comparisons among means for chemical composition, protein spectral band intensity characteristics, secondary structures of protein, initial and final protein digestibility, initial and final releasing percentage of amino acids, and the slope of regression formula were tested using Tukey Honestly Significant Difference (HSD) method. Correlation analysis between protein spectral band intensity characteristics or physicochemical properties and the protein digestibility or the releasing percentage of total amino acids (TAA) was performed using Pearson's correlation coefficient with 95% confidence limits (p < .05). To describe the protein digestibility or releasing percentage of TAA response to hydrolysed time four response models were selected, quadratic, cubic, exponential, and logarithmic. The mathematical expression of these models was as follows.

F ¼ MSR MSE
where MSR is mean square regression.
The best model that was statistically significant, with the largest R 2 , the smallest error residual and RMSE n. The equation for the best model for each feedstuff was used in economic analysis to predict the maximum Yª for x as an interval time: which, was obtained by equating the first derivatives of the response equations to zero, solving for X, for the quadratic and cubic models.
The multiple linear regression analysis for the final protein digestibility or the releasing percentage of TAA based on physicochemical properties was performed using the regression model of . Statements of statistical significance were based on p .05. The data are presented as means ± standard deviation (SD).

Protein solubility
The data of protein solubility % among different feedstuffs is depicted in Figure 1. In general, the soybeans and their products showed the higher protein solubility % (more than 50%), while DDGS, corn germ meal, corn gluten meal, and feather powder had the lower solubility % in descending order. Among ten feedstuffs, the fermented soybean meal exhibited higher protein solubility than other feed ingredients (p < .001), whereas feather meal had lower protein solubility compared to the other feed ingredients (p < .001).

Protein spectral band intensities and protein secondary structures
The intensity characterisation of molecular spectral bands in different protein-rich feed ingredients is depicted in Table 2. The fermented soybean meal, dehulled soybean meal, soybean meal and fish meal had the lower values for amide I-to-II height and area ratios, whereas, the DDGS, corn germ meal and feather meal had the higher values for amide I-to-II height and area ratios (p < .001).
Data presented in Table 3 indicates that the soybean meal, rapeseed meal, fermented soybean meal and the dehulled soybean meal had higher percentages of a-helix structures in descending order compared to all the other feedstuffs (p < .001), While the lower percentages of a-helix were found in corn gluten meal, corn germ meal and feather meal compared to the other feedstuffs (p < .001). The higher values of b-sheet were found in the corn gluten meal and feather meal compared to the other feedstuffs (p < .001), while the fermented soybean meal had a lower value compared to the other feedstuffs (p < .001).
The higher values of b-turn were observed in the soybean meal, fermented soybean meal, corn gluten meal, corn germ meal in descending order compared to the other feedstuffs (p < .001), while the lowest value was observed in the soybean compared to the other feedstuffs (p < .001).
The highest value of random coil was observed in the soybean, while the lower values were observed in the fermented soybean meal and corn gluten meal compared to the other feedstuffs (p < .001). The fermented soybean meal had the highest ratio of a-helix/b-sheet compared to the other feedstuffs (p < .001), while the corn gluten meal had the lower value of such ratio compared to the other feedstuffs (p < .001). Protein digestibility in vitro and in vitro releasing of amino acid Table 4 shows the protein digestibility during trypsin hydrolysis from 0 h to 28 h among ten protein feedstuffs. The initial and final protein digestibility differed in ten protein feedstuffs (p < .001). The fermented soybean meal had a higher percentage of both initial and final protein digestibility than the other feedstuffs (p < .001), while the feather meal had a lower percentage of both initial and final protein digestibility than the other feedstuffs (p < .001). The final protein digestibility of soybean meal, dehulled soybean meal, fish meal, and rapeseed meal was more than 50%, while DDGS, corn germ meal, corn gluten meal, and feather meal were below 50%. Three fitted models between the time (x) and the accumulated digestibility (y) for each feedstuff were highly significant for all proteinrich feedstuffs. regarding the goodness of fit exponential and logarithmic model have approximately value for R 2 . While exponential model has smallest value for RMSEn (0.1%:0.6%) <10% so this model is excellent. Thus, the Exponential model was the best model to describe the relationship between the accumulated digestibility (y) and time (x) Table (S1). The same trend was observed for AA Accumulated releasing percentage   Note: Means with different letters in the same column are significantly different (n ¼ 3, p < .05). a a-Helix is formed by the rotation of the a-carbon atoms of the amino acids in the peptide planes of the protein molecule so that the atoms of the polypeptide main chain are coiled to the right along the central axis. b b-sheet refers to the extended zigzag folding conformation of polypeptide chain with peptide unit as unit and Ca as rotation point. c b-Turn is the structure in which a 180 turn occurs in the polypeptide chain. d Random coil is an irregular conformation of random arrangement of peptide planes in a polypeptide chain. DDGS: dried grains with solubles. percentage, the cubic model was the best model for all feedstuffs Table (S4).
Regarding the values of protein digestibility among ten feed types, all soybean sources and rapeseed meal had the higher regression coefficient and therefore had greater solubility than the other feedstuffs (Table 5). Regarding, the maxim calculated values from the best model of protein digestion against time for each type of feed, rapeseed meal, fish meal and all soybean sources had the higher values and shorter times (the third and fourth period), while the other feedstuffs had their higher values in the fifth period (Table 7).
The in vitro release of TAA during trypsin hydrolysis from 0 h to 28 h presented significant differences among ten protein feedstuffs as shown in Table 8. The lower coefficient value of releasing coefficient of amino acids was observed in DDGS, corn germ, corn gluten meal and feather meal ( Table 6). The same feed types had a maximum calculated value at the fifth period (Table 9). Data presented in Figure 2 represents six consecutive intervals for protein digestibility of different rich-protein feedstuffs. The earlier protein-digestion peak was found in the fermented soybean meal at the third interval time (12-16 h), compared to the other feedstuffs (p < .001). While, the later protein-digestion peaks were observed at the 5 th time interval (20-24 h) in the corn germ meal, DDGS meal, corn gluten meal, and feather meal compared to the other feedstuffs (p < .001). The protein digestibility dropped largely between time intervals 5-6, which indicated the digestibility per unit time decreased owing to the closer maximum of accumulated protein digestibility. In Table 10, the accumulated releasing coefficient of the essential amino acids (EAAs) of different feedstuffs had also the same trend as the TAA. The exact same trend was observed for the release of the TAA (Figure 3).

Correlation analysis between physicochemical properties and the final in vitro protein digestibility or the final in vitro releasing coefficient of TAA
The correlation coefficient (r) between protein spectral band intensities (height and area) and the final in vitro protein digestibility or the final in vitro releasing coefficient of TAA is shown in Table 11. The ratio of amide  I to II height was negatively correlated with the final protein digestibility or the final releasing coefficient of TAA (p < .05). There was a negative correlation between the ratio of amide I to II area and the final protein digestibility or the final releasing coefficient of TAA (p < .05).
The correlation analysis between the secondary protein structures, or solubility and the final protein digestibility or the final releasing coefficient of TAA is depicted in Table 12. The protein digestibility or the final releasing coefficient of TAA had positive correlations with a-helix, a-helix/b-sheet and solubility (p < .05). However, there was a negative relationship between the b-sheet structure and protein digestibility or the final releasing coefficient of TAA (p < .05).

Multiple linear regression analysis of final in vitro protein digestibility and final in vitro releasing coefficient of TAA based on physicochemical properties
Based on the results of correlation coefficients, three physicochemical factors were selected for the regression analysis. The multiple linear regressions between the in vitro protein digestibility or the in vitro releasing coefficient of TAA and physicochemical properties (b-sheet, a-helix/b-sheet and solubility) were obtained (Table 13). It was based on b-sheet (x 1 ), a-helix/b-sheet (x 2 ) and protein solubility (x 3 ) as the independent variable and the protein digestibility or the releasing coefficient of TAA as the dependent variable (y). The multiple correlation coefficients (r 2 ) of all the regression equations were over 0.87.

Discussion
In order to increase protein utilisation and efficiency by optimising protein sources, it is necessary to Table 7. Coefficient value, standard error and regression equation for the best model for protein digestibility coefficient per unit time of feedstuffs and the estimated hydrolysis time which provides the maximum protein digestion rate. Cubic y ¼ 0.047 À 0.035 x þ 0.015 Â 2 À 0.002Â 3 4.12 0.034 a Regression was performed based on the order of unit time (1,2,3 … ) as the independent variable (x) and the protein digestibility coefficient per unit time as the dependent variable (y) during trypsin hydrolysis. b The order of unit time during hydrolysis. c Maximum varied protein digestibility coefficient per unit time. DDGS: dried grains with solubles. 0.014 ± 0.001g 0.243 ± 0.003i p-value <0.001 <0.001 a Note: Means with different letters in the same column are significantly different (n ¼ 3, p < .05). The initial and final releasing efficient was the digestibility after 0 h and 28 h hydrolysis with trypsin, respectively. DDGS: dried grains with solubles; TAA: total amino acids. maximise our knowledge of the nutritional value of protein-rich feed ingredients. The protein digestibility represents an important indicator to evaluate the nutritional value and possesses a close relation with their protein structure and solubility (Yu 2010;Samadi et al. 2013;Peng et al. 2014;Yan et al. 2014;Bai et al. 2015). The releasing coefficient of amino acids during digestion is another important indicator to evaluate Table 9. Coefficient value, standard error and regression equation for the best model for AA releasing coefficient per unit time of feedstuff and the estimated hydrolysis time which provides the maximum AA releasing rate.  Figure 2. Protein digestibility per unit time from different feedstuffs. Note: The abscissa 1, 2, 3, 4, 5, 6 represent from 4 to 8h, from 8 to 12h, from 12 to 16h, from 16 to 20h, from 20 to 24h, and from 24 to 28h, respectively. A dot with an elliptical circle represents the digestibility peak compared to the other intervals.
the nutritional value of feedstuffs since the protein is actually in vivo utilised as amino acids. In the herein study, the relationship between the nutritional value of different protein feedstuffs and their physicochemical properties was further determined.
We used the in vitro enzymatic digestion to evaluate the in vitro protein digestibility and the releasing coefficient of amino acid, because it is suitable for rapid batch detection for such kind of analysis (Boisen and Fern aNdez 1995). The EAA was supplied by diets, which could not be synthesised by animals. Whereas the non-essential amino acids (NEAA) could be synthesised by animals, the release of NEAA during digestion does not indicate the utilisation of nitrogen exactly. Thus, the releasing variation of NEAA wasn't summarised in this study. In this study, the protein digestibility of soybean meal dehulled soybean meal, and rapeseed meal was lower than that of the previous study of Boisen and Fern aNdez (1995), whereas the digestibility of fish meal was higher than the Note: Means with different letters in the same column are significantly different (n ¼ 3, p < .05). k means that the slope value in the accumulated releasing coefficient of EAA was performed based on the regression between the hydrolysed time (h) and the accumulated releasing coefficient. values reported by Huang et al. (2000). This variation might be related to the feed sources, processing technology, or storage conditions (Peng et al. 2014). Whereas, the current finding is consistent with Bao et al. (2014); Jeong et al. (2016) who stated that the processed products of soybean had higher protein digestibility than intact soybeans, which can be explained by the destruction of macromolecular proteins, the deactivation of anti-nutritional factors, or both actions. In the present study, the TAA and EAA of fermented soybean meal had the rapid release and in a similar trend to the protein digestibility. These findings are related to two reasons. Firstly, the rich free amino acids and small peptides in fermented soybean meal can promote preferential hydrolysation in the gastrointestinal tract (Bao et al. 2014). Secondly, the anti-nutritional factors in soybean are eliminated by the action of microbial fermentation, which changes the physicochemical properties of the whole protein molecules, and consequently facilitates the releasing rate of amino acids (Bao et al. 2014). Our results indicated that feather meal and corn gluten meal had a slower rate of digestion and amino acids releasing throughout the in vitro digestion. This result may be attributed to the poor water solubility of feather meal and corn gluten, since the feather meal possesses high content of keratin, and corn gluten meal possesses high content of gliadin that are related to a large amount of hydrophobic amino acids (Papadopoulos et al. 1986). Although soybean and rapeseed meal contain abundant proteins, the antinutritional factors lead to low protein utilisation (Huang et al. 2010;Wang et al. 2019). Rapeseed meal is also rich in crude fibre, which may negatively affect the digestibility of nutrients (mainly proteins, amino acids, and minerals). The protein solubility (as an important functional property) is an indicator of potential application value related to protein digestibility (Pengbin et al. 2002;Bai et al. 2015). The solubility of fish meal, DDGS and corn gluten meal was similar to the previous result of Bai et al. (2015), while soybean meal protein in this study was lower than the same previous study, which may be attributable to the original source (Bai et al. 2015). Besides protein digestibility, we found that the releasing coefficient of amino acids also showed a positive correlation with protein solubility, which means that the higher protein solubility value is related to the high releasing coefficient of amino acids.
Fourier transform infra-red (FTIR) spectroscopy is an important tool to study the protein conformation (Tatulian 2019). Amides I and II are the most useful bands that can reveal changes in the main structures of protein (Wei et al. 2018 Figure 3. Releasing coefficient of TAA per unit time from different feedstuffs. Note: The abscissa 1, 2, 3, 4, 5, 6 represent trypsin digestion time from 4 to 8h, from 8 to 12h, from 12 to 16h, from 16 to 20h, from 20 to 24h, and from 24 to 28h, respectively. The dot with an outside elliptical circle represents that the releasing coefficient of TAA per unit time in this period was higher than that in the other periods. TAA: total amino acids. the protein amide I to amide II ratio in the different protein sources had a strongly negative correlation with protein digestibility or releasing coefficient for the TAA under in vitro conditions. The slow release of DDGS and corn germ meal might be related to the lower peak height and peak area of amide II band. Such lower values of amide II band indicate the lower content of the basic amino acid in these feedstuffs, and also indicate that the peptide cleavage site of pepsin and trypsin is a peptide bond formed by Lys and Arg amino acids (Adler-Nissen 1979). DDGS and corn germ meal presented low basic amino acids (Lys, Arg and His) in this study. Therefore, the relative lack of basic AA may lead to reduce the actions of pepsin and trypsin (Adler-Nissen 1979), and consequently, the pace of amino acids releasing coefficient is decreased.
The proteins contain a variety of different secondary structures. The amide I band is highly sensitive to the secondary structure components of protein , and deconvolving the amide I band is found in a-helix, b-sheet, b-turn, and random coil. Several studies have shown that the nutritional value of protein is affected by protein secondary structures . For example, the ratio of a-helix to b-sheet height is associated with the solubility and protein degradability of plant-derived protein feeds (Yu 2005;Doiron et al. 2009). In this experiment, the a-helix or the ratio of a-helix to b-sheet were positively correlated with in vitro protein digestibility and releasing coefficient of TAA, but the relative mechanism is still unclear. On the contrary, the b-sheet possesses negative correlations with the digestibility and the amino acids releasing coefficient. These results are similar to previous studies of Abeysekara et al. (2011);Yu and Damiran (2011); Khan et al. (2015). b-sheet is rich in hydrogen bonds, which can prevent the digestive enzymatic activities of proteins (Bai et al. 2015).
In the herein study, we estimated the multiple regression model among three of the main physicochemical characteristics (b-sheet, a-helix/b-sheet and solubility) of rich-protein feedstuffs to predict the digestibility or releasing coefficient of TAA of protein source. The a-helix to b-sheet ratio had higher r to predict the protein digestibility or releasing coefficient of TAA, compared to a-helix. This ratio was an effective predictor of protein digestibility of protein feedstuff in the previous studies on canola meal and press cake , and Brassica carinata seed Xin et al. 2013). To some extent, the characterisation of secondary protein structure could predict the variation in protein digestibility and release of amino acids.

Conclusions
In conclusion, during the in vitro digestion, the releasing coefficient of amino acids for dietary protein had a similar trend to protein digestibility. The physiochemical properties of protein feedstuff were closely related to the variation of protein digestibility or releasing coefficient of TAA. The physiochemical properties of dietary protein (such as b-sheet, a-helix/b-sheet and solubility) could be used as predictors for its digestibility or releasing coefficient of TAA.