Structure of native proteins

Proteins are polymers made up of different amino acids linked by peptide bonds. This linear chain of amino acids forms the primary structure of the protein and their sequence, specific for every protein, also determines the secondary, tertiary and quaternary structures of the native form of the protein⁽ Reference Davis and Williams ^{24 )}. The main configurations of the secondary structure of protein are the α-helices and β-sheets⁽ Reference Guan, Marcos and O’Connell ^{25 ,} Reference Berman, Westbrook and Feng ^{26 )} (Fig. 1), although others exist, mostly resulting from variations of these configurations. The α-helices and β-sheets provide strength and rigidity to proteins⁽ Reference Gropper, Smith and Groff ^{27 )}. Regions in the primary structure of a protein that lack structural elements are considered as random coils⁽ Reference Smith, Fiebig and Schwalbe ^{28 )}, which by definition are less stable⁽ Reference Gropper, Smith and Groff ^{27 )}. The secondary structure of proteins can be determined by several techniques, such as Fourier transform IR (FTIR) spectroscopy and circular dichroism, the latter one used solely for soluble proteins. Circular dichroism uses the polarisation angle of polarised light in the far-UV region for the determination of the secondary structure of proteins and the near-UV region for the determination of changes in the tertiary structure. With FTIR spectroscopy, the amide I region (1600–1700/cm) is used to measure the secondary structure of proteins by detecting the vibrations of the carboxyl groups of amino acids⁽ Reference Hu, Kaplan and Cebe ^{29 ,} Reference Carbonaro, Maselli and Nucara ^{30 )}. This region of the spectra includes bands for the main secondary structures: α-helices (1650–1660/cm) and β-sheets (1630–1638/cm). Within the amide I region, also bands that indicate modifications to the native structural conformation can be distinguished, such as the intermolecular β-sheets (1620–1630/cm) and A2 bands (1690–1695/cm), indicative for protein aggregation⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}. Intermolecular β-sheet structures originate from intermolecular hydrogen bonds⁽ Reference Shivu, Seshadri and Li ^{31 )} and A2 regions from intermolecular aggregation of proteins or the vibration of the carboxyl groups of the amino acid side chains⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}.

Fig. 1 Cartoon view of the main secondary structures of proteins. The protein in the cartoon is Protein Data Bank ID: 4R80⁽ Reference Guan, Marcos and O’Connell ^{25 ,} Reference Berman, Westbrook and Feng ^{26 )}.

Intramolecular formation of non-covalent and disulfide bonds between side chains of amino acids holds the three-dimensional tertiary structure of protein together. The hydrophobic amino acids are hidden in the interior of the structure. Non-covalent bonds between amino acids can be due to Van der Waals interactions between hydrophobic amino acids (for example, proline, tryptophan), to hydrogen bonding, or to electrostatic interactions between amino acids with opposite charges (for example, lysine and glutamate)⁽ Reference Gropper, Smith and Groff ^{27 )}. Moreover, proteins in their tertiary structure can interact with each other and form non-covalent and covalent (disulfide) intermolecular bonds, which give rise to the quaternary structure of proteins. Most of the storage proteins in vegetable sources are present in a quaternary conformation, which can also be considered their native conformation. Examples of such proteins are napin (albumin) and cruciferin (globulin) in rapeseed, and the globulins β-conglycinin and glycinin in soyabean. Rupture of intermolecular bonds releases the individual proteins that compose the quaternary structure. For example, under reducing conditions, the acidic and basic polypeptides of soyabean glycinin and rapeseed cruciferin are separated as seen by electrophoresis (SDS-PAGE)⁽ Reference Inquello, Raymond and Azanza ^{32 ,} Reference Hou and Chang ^{33 )}. The energy required to break disulfide covalent bonds is higher than the energy required to break non-covalent bonds (Table 1). Proteins with large numbers of disulfide bonds are regarded to have a higher resistance to enzymic activity during digestion and also a lower sensitivity to structural changes during processing than proteins with small numbers of disulfide bonds⁽ Reference Clemente, Vioque and Sánchez-Vioque ^{34 )}.

Proteins from cereals and legumes are usually stored as protein bodies. In general, approximately 80 % of the protein content of legumes can be classified as storage proteins and is stored in protein bodies⁽ Reference Sikorski ^{35 )}. It has been defined that any protein present at least in 5 % of the total protein content can be considered as storage protein⁽ Reference Derbyshire, Wright and Boulter ^{36 )}. The remainder of the proteins, which comprise the proteins with biological activity (for example, enzymes and inhibitors), are not contained within protein bodies.

An overview of the structural properties of the major proteins that are present in important vegetable ingredients used for commercial swine compound feed production is shown in Table 2. The proteins are described based on the most abundant protein types for each ingredient. For example, although classified under the same name (i.e. zein), the prolamins from maize consist of four types of proteins with different secondary structures (i.e. α-, β-, γ- and δ-zein). The secondary structure of α-zein is mostly α-helical, whilst that of β-zein (which is also related to γ-zein) consists mostly of β-sheets and turns⁽ Reference Kim, Woo and Clore ^{37 )}. The difference in the secondary structure of these proteins originates from their different amino acid compositions⁽ Reference Shewry and Tatham ^{38 )}.

Overall, globulins are the predominant protein types in oil (for example, rapeseed) and legume seeds (for example, soyabean), whilst prolamins and glutelins are the most abundant cereal proteins. In addition, proteins with very different structural characteristics can be found within the same ingredient. For example, cruciferin in rapeseed has a higher molecular weight and α-helix:β-sheet ratio than napin. According to the ‘Osborne’⁽ Reference Osborne ^{39 )} classification of proteins, albumins are soluble in water, globulins in diluted salt solutions, prolamins in ethanol–water solutions and glutelins in diluted alkali solutions. Although many proteins are classified as being from the same ‘Osborne’ type⁽ Reference Osborne ^{39 )}, proteins from different ingredients usually differ in their amino acid composition. This is reflected in different conformations, thereby making it impossible to generalise on the types of bonds present and their secondary, tertiary or quaternary structures. Globulins and albumins seem to have a higher content of intramolecular β-sheet structures compared with prolamins. This is also reflected in a lower α-helix:β-sheet ratio of the globulins and albumins compared with the prolamins.

Carbonaro et al. ⁽ Reference Carbonaro, Maselli and Nucara ^{30 )} suggested that the secondary structure of the protein in a (processed) ingredient is a good predictor of digestibility. These authors reported that the proportion of intramolecular β-sheet structures in food sources from vegetable and animal origin is negatively correlated with in vitro crude protein digestibility (r –0·98) using porcine trypsin and peptidase, bovine chymotrypsin and a bacterial protease (the latter added in a subsequent incubation). A higher α-helix:β-sheet ratio has been described to negatively influence in vitro intestinal digestibility of rumen undegraded protein in ruminants⁽ Reference Yu and Nuez-Ortin ^{40 )}, although the mechanism of digestion in ruminants differs from that in single-stomached animals. The relationship between the α-helix:β-sheet ratio and digestibility was also studied in vitro ⁽ Reference Rubio, Pérez and Ruiz ^{41 )}. It was found that the in vitro digestibility of the albumin, globulin and vicilin fractions of peas (41, 63 and 88 %, respectively), using porcine pepsin and pancreatin, corresponded with the α-helix:β-sheet ratio of 0·06, 0·20 and 0·80, respectively. Variation in the secondary structure of proteins, however, does not seem to explain the variation in digestibility completely. Napin and cruciferin, which have very different secondary structures (for example, ratio α-helix:β-sheet is 0·66 in napin and 0·35 in cruciferin), do not differ much in their apparent faecal digestibility in rats⁽ Reference Delisle, Amiot and Goulet ^{42 )}. Also, proteins with similar secondary structures, such as glycinin and β-conglycinin from soyabeans, differ in their apparent faecal digestion coefficients⁽ Reference Delisle, Amiot and Goulet ^{42 )}. The larger number of disulfide bonds in glycinin compared with β-conglycinin has been postulated to provide a further explanation for the differences in the digestibility values between these proteins⁽ Reference Hou and Chang ^{33 ,} Reference Hou and Chang ^{43 )}.

There was no correlation (P>0·05) between the secondary structure of isolated native proteins and in vitro or faecal digestibility (Table 2). However, literature on food allergens⁽ Reference Bannon, Goodman and Leach ^{44 ,} Reference Bannon, Fu and Kimber ^{45 )} suggests that the native structure of some proteins (for example, Ara h 1 from groundnuts or β-conglycinin from soyabeans) influences the accessibility of gastric proteases, which might also limit the access of intestinal peptidases for proteolysis. For example, the main globulin from kidney beans (Phaseolus vulgaris), phaseolin, has high resistance to pepsin and pancreatin hydrolysis (19 % in vitro digestibility), which was attributed to its closed tertiary or quaternary structure that restricts enzyme accessibility⁽ Reference Montoya, Leterme and Victoria ^{46 )}.

Structural changes due to protein denaturation

Exposure of plant storage proteins to harsh conditions, like extreme pH values, heat or pressure, causes protein denaturation. For example, a pH higher than 11·5 or lower than 3·0 or pressure at 400 MPa for 10 min cause complete denaturation of soya glycinin⁽ Reference Kim, Kim and Yang ^{47 ,} Reference Zhang, Li and Tatsumi ^{48 )}. Denaturation is the unfolding of proteins from their tertiary or secondary structures and will occur when the denaturing influence is sufficiently large to break the non-covalent or covalent bonds that hold the structure together. Conditions for denaturation depend on the type and structure of each native protein⁽ Reference Kilara and Sharkasi ^{49 )}. Refolded proteins with a structural conformation different from their native state can also be considered as denatured. Proteins refold in an attempt to minimise their free energy state, with simultaneous formation of new bonds or interactions⁽ Reference Becker and Yu ^{50 )}. It has been suggested⁽ Reference Marsman, Gruppen and Mul ^{51 )} that the new bonds formed are not representative of the original conformation of the protein, as this is statistically unlikely to occur. These new bonds result from intramolecular electrostatic and hydrophobic interactions between amino acids, or from intermolecular non-covalent and/or disulfide bonds between two unfolded proteins. The formation of these bonds or interactions between amino acids may lead to protein aggregates, which can be soluble or insoluble depending on the molecular weight of the aggregate. As defined by Wang et al. ⁽ Reference Wang, Nema and Teagarden ^{52 )}, protein aggregates consist of proteins that have lost their native state and are at least twice the size of the native protein. The mechanisms that explain the aggregation of proteins were extensively reviewed by these authors. Protein aggregation can also occur due to crosslinking reactions between proteins, resulting from protein oxidation and the formation of covalent bonds between amino acids, which is favoured at alkaline pH⁽ Reference Schwarzenbolz and Henle ^{53 )}. An example of the latter is the formation of lysinoalanine, which is a crosslink between lysine and alanine and is mediated through the formation of dehydroalanine from β-substituted amino acids (for example, phosphoserine) and the subsequent nucleophilic addition of the ε-amino group of lysine⁽ Reference Schwarzenbolz and Henle ^{53 )}.

During digestion, proteins are partially denatured by the acidic pH conditions in the stomach, which facilitates the accessibility of proteases to the peptide bonds. Pepsin is secreted in the stomach and cleaves peptide bonds between hydrophobic or neutral amino acids, except for proline. Proteins that are resistant to the acidic and proteolytic conditions in the stomach can reach the intestinal mucosa and cause allergic reactions⁽ Reference Bannon, Goodman and Leach ^{44 )}. Trypsin and chymotrypsin are secreted in the pancreatic juices in the duodenum. Trypsin is highly specific to lysine and arginine, whilst chymotrypsin has a preference for large hydrophobic amino acids, such as tryptophan, tyrosine and phenylalanine. This specificity becomes important when the cleavage sites for these enzymes are physically or chemically blocked by structural constrains, such as protein aggregation due to non-covalent or covalent bonds formation or by modification of the amino acid residues due to protein oxidation (for example, dityrosine bonds) or Maillard-type reactions.

Changes in the structure of proteins due to denaturing treatments, such as heat, pH and pressure, can lead to proteins that become either more susceptible or more resistant to proteolysis. Protein denaturation can result in the formation of random coils, which exposes groups that are not usually accessible in the native form, thus becoming more susceptible to enzymic hydrolysis⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Meade, Reid and Gerrard ^{54 )}. The formation of random coils due to processing results from the breakdown of the tertiary structure and that of the secondary structure, such as α-helices and intramolecular β-sheets. Denaturation, especially at extreme conditions involving high temperatures or pH extremes, promotes protein aggregation⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Carbonaro, Cappelloni and Nicoli ^{55 –} Reference Tang, Chen and Ma ^{59 )} and crosslinking between amino acids⁽ Reference Gerrard ^{60 )}. Aggregation and crosslinking reactions could decrease the accessibility of digestive enzymes, thereby reducing protein digestibility⁽ Reference Pinto, Léonil and Henry ^{61 ,} Reference Rérat, Calmes and Vaissade ^{62 )}.

The extent of structural changes in proteins depends on the conditions employed during processing, and also on the types of proteins present. Upon thermal processing of rapeseed protein isolates, the globulin fraction (cruciferin) was more affected and formed a larger amount of aggregates than the albumin fraction (napin)⁽ Reference He, He and Chao ^{63 )}. Similar results have been reported for soyabean proteins, in which β-conglycinin is more heat-sensitive than glycinin due to the disulfide bonds present in the latter protein type⁽ Reference Marsman, Gruppen and Mul ^{51 )}. Disulfide bonds can be broken during processing, but depending on the severity of the conditions, the free thiol residues can also react to form new bonds via sulfydryl-disulfide interchange reactions and increase aggregation⁽ Reference Tang, Chen and Ma ^{59 )}. Differences in amino acid composition and structure between different types of the same protein (for example, α-, β- and γ-kafirins) might also influence the response of the proteins to denaturing conditions and digestion⁽ Reference Oria, Hamaker and Shull ^{64 )}.

Furthermore, it has been suggested that the net charge of the protein influences the structural response to thermal processing⁽ Reference Tang, Chen and Ma ^{59 )}. Increasing NaCl concentrations from 0 to 1·5 m at pH 8 increases the denaturation temperature of soya glycinin from 79·1 to 98·5°C⁽ Reference Kim, Kim and Yang ^{47 )}. This is an indication of a protective effect of the salt on the native structure of glycinin from protein denaturation at increasing temperatures. When similar processing conditions were used, a higher content of hydrophobic and uncharged polar amino acids and higher acidic:basic amino acid ratio were linked to a decrease in aggregation⁽ Reference Tang, Chen and Ma ^{59 )}. It has been proposed⁽ Reference Carbonaro, Cappelloni and Nicoli ^{55 )} that heat-induced aggregation promoted by basic amino acids (lysine, arginine or histidine) could have detrimental effects on protein digestibility.

In addition, proteins are highly susceptible to react with reducing sugars upon heating to produce Maillard reaction products. Some essential amino acid residues, such as lysine and arginine, are more susceptible than other amino acids to these type of reactions. Lysine susceptibility is due to the presence of the additional amino group, whilst in arginine it is due to the additional guanidinium group⁽ Reference van Rooijen, Bosch and van der Poel ^{65 )}. The nature of these reactions and the wide range of compounds formed have been reviewed elsewhere⁽ Reference van Rooijen, Bosch and van der Poel ^{65 )}. It must be considered that structural changes per se (for example, protein aggregation) and chemical changes due to Maillard reactions may occur simultaneously⁽ Reference Gerrard, Lasse and Cottam ^{66 )} and that the effects of both on proteolysis might be confounded.

Protein digestion is a multifactorial process that is influenced not only by the physico-chemical properties of the proteins, but also by the matrix in which the protein is embedded. Microstructural components of the feed are not often described in the literature and this aspect should be taken into account in future research that links modifications of proteins due to processing to protein digestibility.

Heat-induced changes

Thermal treatments are involved in most common processing technologies to achieve the desired physical state of the ingredients and ingredient mixtures. Other reasons are to improve the digestibility of starch and proteins, to eliminate residual organic solvents after oil extraction and antinutritional factors, and to maintain hygienic levels of the processed feeds⁽ Reference Abdollahi, Ravindran and Svihus ^{15 ,} Reference Abdollahi, Ravindran and Wester ^{70 –} Reference Van der Poel ^{72 )}.

The maximum temperature achieved during commercial oil extraction of soyabeans and rapeseeds ranges between 100 and 110°C⁽ Reference Classen, Newkirk and Maenz ^{10 ,} Reference Paraíso, Cauneto and Zemp ^{11 ,} Reference Newkirk and Classen ^{73 ,} Reference Pope, Karr-Lilienthal and Utterback ^{74 )}, a temperature which is held constant for 60–90 min⁽ Reference Newkirk, Classen and Edney ^{12 )}. It must be emphasised that these temperatures are achieved by toasting, which comprises the use of steam, which is a source of heat that involves large amounts of moisture. Large differences with respect to the formation of Maillard reaction products in soyabean meal proteins have been detected between dry (for example, oven-drying) and wet (for example, autoclaving) sources of heat⁽ Reference Gonzalez-Vega, Kim and Htoo ^{75 )}, making it difficult to disentangle the individual effects of heat and moisture.

The temperature applied during compound feed production (secondary processing) depends on the technology employed. Common pelleting temperatures for swine feed production range between 60 and 100°C, whilst higher temperatures are achieved with expander processing (90–130°C) and extrusion (60–160°C)⁽ Reference Chae and Han ^{76 )}.

Changes in the secondary structural conformation of proteins due to thermal processing have been reported (Table 3), although only a limited number of studies have linked these changes to protein digestibility. After autoclaving at 120°C for 20 min, an increase in the relative amount of random coil conformation in different whole ingredients was reported⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}. Heating at 100°C for 15 min also increased the proportion of random coils in rapeseed protein isolate from 23 % in the unheated material to 28 % after heating, whilst increasing simultaneously the proportion of β-sheets from 10 to 33 %⁽ Reference He, He and Chao ^{63 )}. Formation of random coils is an indication of successful protein denaturation, also linked to a positive correlation with in vitro digestibility (r 0·91)⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}. The increase of the in vitro digestibility reported in this study after processing ranges from 2 to 6 %, which can be considered as low. In addition, inactivation of protease inhibitors was not considered in this study. The effect of the formation of random coils on the increase in digestibility after the thermal treatment might be confounded with the inactivation of protease inhibitors. These authors also reported that autoclaving induced the disappearance of the intramolecular β-sheet structures and the appearance of intermolecular β-sheet structures and A2 bands in FTIR spectra. These two bands in the spectra (intermolecular β-sheets and A2 bands) have been related to protein aggregation and the proportion of intermolecular β-sheets was highly predictive (r –0·99) for in vitro protein digestibility. As formation of aggregates and random coils occur simultaneously during processing, the net effect on digestibility is related to the relative frequency of each phenomenon. When taken together (Fig. 2), the total content of intramolecular β-sheet structures from native proteins and the intermolecular β-sheet structures from proteins after thermal treatment has a high negative correlation (r –0·95; P<0·001; n 8) with in vitro crude protein digestibility. The total content of intramolecular and intermolecular β-sheets can be used as a fast predictor for protein digestibility, at least for legume protein sources. As both β-sheet structures within the protein and between proteins correlate well with in vitro crude protein digestibility, it appears that the structure itself (i.e. β-sheets), but not its location, limits proteolysis. The correlation between the α-helix:β-sheet ratio and in vitro crude protein digestibility was not significant, but exhibits a trend (r 0·65; P=0·08). Nevertheless, native legume ingredients tend to have a lower ratio and digestibility compared with thermally treated ingredients. It should be noted, however, that these correlations originate from a single study⁽ Reference Carbonaro, Maselli and Nucara ^{30 )} and that the ingredients used probably contained protease inhibitors. It is likely that these protease inhibitors were inactivated during thermal treatment and that the increase in digestibility can be partly attributed to the inactivation of protease inhibitors and partly to the modifications of the secondary structure.

Fig. 2 Correlation between the β-sheets content and in vitro crude protein digestibility coefficient of native (- - -) and thermally treated (––) whole legume seeds: ○, Phaseolus vulgaris; ●, Cicer arietinum; ∆, Lens culinaris; ▲, Glycine max. Adapted from Carbonaro et al. ⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}.

Although the in vitro digestibility of the most frequent protein fraction present in kidney beans (phaseolin) was improved after thermal processing (Table 4), the digestibility of other protein fractions from the same source (total albumins, protease inhibitor and lectins) was largely reduced⁽ Reference Genovese and Lajolo ^{77 )}. This could explain why minor or no changes in digestibility were reported after thermal treatment of whole kidney beans or protein isolates thereof ⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Carbonaro, Cappelloni and Nicoli ^{55 ,} Reference Tang, Chen and Ma ^{59 )}. The overall effect on total protein in the feedstuff seems to be a combination of the positive and negative effects on each individual protein fraction. This means that for a complete understanding of the underlying mechanisms that explain the effects of processing on protein digestibility in feed ingredients, the effects on each individual protein type needs to be studied. Positive effects of the thermal treatment of kidney beans on apparent ileal crude protein digestibility in pigs have been reported⁽ Reference Van der Poel, Blonk and Huisman ^{78 )}. It was discussed that these effects might be due to inactivation of antinutritional factors and conformational changes of proteins during ‘high temperature short time’ treatments (136°C; 1·5 min), which is in contrast to the lower temperature and longer times (ranging from 95 to 121°C; 15 to 30 min) applied in other studies⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Carbonaro, Cappelloni and Nicoli ^{55 ,} Reference Tang, Chen and Ma ^{59 ,} Reference Genovese and Lajolo ^{77 )}.

Changes in the solubility of proteins, due to the heat-induced formation of insoluble aggregates, have been reported for a wide range of experimental conditions and were linked to changes in the types of bonds within and between proteins (see online Supplementary Table S1). There is a parallel increase in the amount of insoluble protein aggregates with the severity of the thermal treatments applied. Already at 80°C, 31 % of the protein content in rapeseed protein isolate formed insoluble aggregates, compared with 0 % in the unheated material⁽ Reference He, He and Chao ^{63 )}. In the isolate, cruciferin (globulin) was more susceptible to aggregate formation than napin (albumin)⁽ Reference He, He and Chao ^{63 )}. Most studies agree on a decreased solubility of a protein after a thermal treatment, compared with that of the native protein⁽ Reference Li and Lee ^{57 ,} Reference Lin, Huff and Hsieh ^{79 –} Reference Liu and Hsieh ^{81 )}. The reduction in protein solubility is an indication of the formation of insoluble protein aggregates. Many native proteins are soluble as their hydrophobic groups are located on the inside of the molecule. After denaturation, involving reorientation of the hydrophobic groups and refolding, the initial protein aggregates formed are soluble, but they become insoluble when their size exceeds the solubility limit⁽ Reference Shivu, Seshadri and Li ^{31 )}.

Protein solubility studies with agents capable of cleaving different types of bonds (for example, urea, SDS and dithiothreitol (DTT)) have been used in order to study the main causes for protein insolubility (see online Supplementary Table S1). Urea and SDS are good agents to cleave non-covalent bonds, whilst DTT is adequate for the reduction of disulfide bonds into thiols. The solubility of unheated proteins in buffer solutions is similar to the solubility of aggregated proteins with solutions that contain both non-covalent and covalent cleaving agents. Combining these two types of cleaving agents has a synergistic effect on the amount of solubilised protein, compared with the amount of protein solubilised by using them separately⁽ Reference Li and Lee ^{57 ,} Reference Lin, Huff and Hsieh ^{79 –} Reference Liu and Hsieh ^{81 )}. This indicates that non-covalent and covalent bonds are mutually important in maintaining the structure of aggregated proteins.

After thermal processing, there is an increase in the disulfide bond content with a simultaneous reduction in the content of thiols (Table 5). For example, thiol content was reduced from 6·6 μmol/g protein in the albumin and the protein inhibitors plus lectins fractions isolated from native kidney beans to 0·2 and 0·6 μmol/g protein, respectively, after heating at 121°C for 15 min⁽ Reference Genovese and Lajolo ^{77 )}. At the same time, the disulfide bond content was increased from 24·8 to 28 μmol/g protein in the albumin fraction and from 22·5 to 25·5 μmol/g protein in the protein inhibitors plus lectins fraction. These changes were more evident with autoclaving at 121°C for 15 min than with cooking at 99°C for 30 min. This could be an indication that the effect of temperature on protein denaturation is more relevant than the effect of time, as reported elsewhere⁽ Reference Van der Poel, Blonk and Huisman ^{78 )} for the storage protein phaseolin. There was a positive correlation between the in vitro crude protein digestibility and the content of thiol groups in thermally treated isolated albumins from kidney beans (r 0·80; P<0·003; n 11) (Fig. 3). Correlations for the rest of the protein sources analysed were not significant. Some of these sources (Phaseolus beans and soyabeans) probably contained protease inhibitors and lectins. An increase in digestibility that resulted from heat inactivation of protease inhibitors and lectins could mask the decrease in digestibility resulting from heat-induced disulfide bonding. It has been suggested⁽ Reference Selle, Liu and Cai ^{82 ,} Reference Selle, Liu and Cai ^{83 )} that the formation of disulfide bonds after heat treatment of sorghum is responsible for the decreased digestibility of kafirin. However, correlations between disulfide bond content of sorghum after thermal treatments and standardised ileal digestibility were not significant (P>0·05). Oddly, the disulfide bond content in the study of Selle et al. ⁽ Reference Selle, Liu and Cai ^{83 )} did not increase along with the decrease in the thiol content. With increasing temperatures, there is also an increase in the surface hydrophobicity of the proteins in rapeseed protein isolate, which is a reflection of the exposure to the surface of the hydrophobic groups buried inside the molecules⁽ Reference He, He and Chao ^{63 )}. Surface hydrophobicity (S _O ) increased from 600 in the unheated rapeseed protein isolate to 650, 1500 and 1100 after thermal treatments at 60, 80 and 100°C for 15 min, respectively⁽ Reference He, He and Chao ^{63 )}.

Fig. 3 Correlation between the content of disulfide (○) or thiol (●) groups and in vitro protein digestibility coefficient for thermally treated albumins isolated from kidney beans.

Overall changes in the structure of protein can be estimated by comparing the denaturation enthalpy of native protein with that of the protein after processing (see online Supplementary Table S2). Denaturation enthalpy is a measure of the amount of energy required to denature a protein. Hence, denatured proteins exhibit a lower or a lack of denaturation enthalpy compared with the native ones. Protein denaturation usually increases with the intensity of the heat treatment and has been suggested as the main reason for improved protein digestibility in thermally heat-treated ingredients⁽ Reference Abdollahi, Ravindran and Wester ^{70 ,} Reference Mauron ^{84 )}. According to these authors, unfolding of the native secondary/tertiary structure of the protein could facilitate enzymic attack and their suggestion was based on several proteins of vegetable origin (for example, wheat and soya). The conditions applied during processing probably determine whether proteins become more susceptible (for example, unfolding and random coil formation) or more resistant (for example, protein aggregation and chemical changes to amino acids) to enzymic attack. However, enthalpy of denaturation can also increase after long thermal treatments as reported⁽ Reference Wally-Vallim, Vanier and da Rosa Zavareze ^{85 )} for soyabean protein isolates prepared from heat-treated soyabeans at 40 and 80°C for 4, 8, 12 and 16 h. The denaturation enthalpy of this material decreased after 4 and 8 h of thermal treatment, but increased again with longer heating times. It is possible that protein refolding occurred at longer heating times, causing the denaturation enthalpy to increase. Most of the denaturation enthalpy observed, ranging from 36 to 70 % of the enthalpy in the original material (2·85 J/g), originated from glycinin, which was less denatured than β-conglycinin under these conditions.

Only a limited number of studies has been performed on the link between denaturation enthalpy and digestibility of processed vegetable protein sources⁽ Reference Wally-Vallim, Vanier and da Rosa Zavareze ^{85 –} Reference Sun, Mu and Sun ^{88 )}. The positive correlation between the relative degree of denaturation (i.e. decrease of enthalpy after thermal treatment with respect to the enthalpy of the native material) and in vitro crude protein digestibility for protein-containing ingredients (i.e. sweet potato protein, cowpeas and soyabeans) was significant (r 0·49; P=0·003; n 34) (Fig. 4). However, these ingredients probably contained protease inhibitors. In that case, the relationship between the degree of protein denaturation and the in vitro crude protein digestibility would be somewhat confounded with increased apparent digestibility due to the heat inactivation of protease inhibitors. The amino acid sequence of sporamin (sweet potato protein) resembles that of Kunitz-type trypsin inhibitors, possibly explaining the increase in digestibility after denaturation⁽ Reference Sun, Mu and Sun ^{88 )}. Proteins from cowpeas and soyabeans can be classified under the same ‘Osborne’ type. However, their amino acid compositions are different, which possibly leads to different structural conformations (for example, secondary and tertiary) and explains the differences in the correlations between the relative degree of denaturation and in vitro digestibility (Fig. 4). In addition, whilst the in vitro digestibility of sweet potato protein and cowpeas was determined using multiple enzymes (for example, porcine pepsin and pancreatin), the digestibility of soyabean proteins was determined after incubation with only pepsin. Incubation with a single enzyme might lead to an underestimation of protein digestibility. For isolated albumins from P. vulgaris there is a trend for a negative correlation between the relative degree of denaturation and the in vitro digestibility (r –0·53; P=0·09; n 11) (Fig. 4). The degree of purity of the isolated albumin fraction is not described in that study⁽ Reference Rocha, Genovese and Lajolo ^{87 )}; thus this fraction might still contain trypsin inhibitors. The limited number of studies available, the inclusion of sources that contain protease inhibitors and the inconsistency in the results do not allow us to draw firm conclusions on the effects of protein denaturation on protein digestibility. The effect of protein denaturation on protein digestibility could be linked to the nature of the protein.

Fig. 4 Correlation between the relative degree of denaturation and in vitro crude protein digestibility coefficient of thermally treated isolated albumins from kidney beans (◊) and other protein sources (■, Ipomoea batatus; ▲, Vigna unguiculata; ●, Glycine max).

Effects of moisture

Water or steam is usually added as cofactors during processing in order to manage the moisture content and the physical properties of the processed material. The moisture content ranges between 12 and 18 % (w/w) for pelleting⁽ Reference Abdollahi, Ravindran and Svihus ^{15 ,} Reference Chae and Han ^{76 )}, whereas extrusion processing tolerates higher moisture inclusion ranging from 20 to 35 % (w/w)⁽ Reference Chae and Han ^{76 )}.

Proteins in dry conditions have high glass transition temperatures (T_g), which can be close to their degradation temperature⁽ Reference Verbeek and van den Berg ^{89 ,} Reference Bier, Verbeek and Lay ^{90 )}. The T_g can be reduced by the addition of water, which acts as a protein plasticiser, and decreases T_g approximately 10°C for every 1 % increase in moisture content⁽ Reference Bier, Verbeek and Lay ^{90 )}. Due to the small size of its molecules, water can move freely through the small openings of the structure of proteins. Water decreases the T_g by disrupting hydrogen bonds, van der Waals and ionic interactions that hold the structure of protein together, thereby increasing its flexibility⁽ Reference Verbeek and van den Berg ^{89 )}. Increased flexibility and rupture of the bonds also decrease the denaturation temperature of proteins. However, the flexibility of proteins is limited below water contents of 5–9 %. Exposure of hydrophobic amino acid residues due to changes in the conformation of proteins after the addition of water can lead to irreversible protein aggregation⁽ Reference Towns ^{91 )}. The denaturation temperature of sunflower (Helianthus annuus) oil cake globulins decreased from 189·5°C to 154·4, 133·4 and 119·9°C with increasing moisture contents of 0, 10, 20 and 30 %, respectively⁽ Reference Rouilly, Orliac and Silvestre ^{92 )}. The T_g of these proteins showed a similar pattern, decreasing from 180·8°C with no water to 5·3°C at 26 % moisture content. Similar results were also reported for β-conglycinin⁽ Reference Kitabatake and Doi ^{93 )} and glycinin⁽ Reference Kitabatake and Doi ^{93 ,} Reference Huson, Strounina and Kealley ^{94 )} from soyabean meal. The denaturation temperature of β-conglycinin and glycinin were 76·5°C and 93·3°C, respectively, with a moisture content of 94 %. However, when the moisture content was decreased to 29 %, the denaturation temperature of β-conglycinin was increased to 180°C and no denaturation temperature could be found for glycinin. The addition of water decreases the T_g of both native and denatured glycinin, with a faster rate for the native proteins⁽ Reference Huson, Strounina and Kealley ^{94 )}.

Thus, water addition increases the flexibility of protein structure and in combination with heat, shear or pressure, changes the effects of these factors on the secondary protein structure (Table 3), in vitro crude protein digestibility (Table 4) and protein solubility (see online Supplementary Table S1). Increasing the moisture content during extrusion processing, whilst keeping a constant temperature, increases protein denaturation and refolding, with the formation of new bonds. For example, the percentage of protein solubilised by a solution of 50 mm-DTT in a phosphate buffer after extrusion at 170°C of a soya protein, wheat gluten and wheat starch mix (weight ratio 60:40:5) increased from 10 to 30 when moisture content (w/w) was increased from 60 to 72 %⁽ Reference Liu and Hsieh ^{81 )}. This indicates that during processing the extents of both covalent and non-covalent interactions increase with increasing moisture contents, possibly related to unfolding, which is the rate-limiting step. The increased interaction between proteins could explain the reduction of the in vitro digestibility of soyabean meal with increasing moisture contents at constant temperature and extrusion screw speed (Table 4). Alternatively, inactivation of trypsin inhibitors could have been higher at lower moisture contents, leading to higher in vitro crude protein digestibility.

Changes in the secondary structure of proteins have been reported when moisture-including sources of heat were used (for example, using autoclaving, desolventisation/toasting) in contrast with dry sources of heat (for example, roasting, IR radiation) at similar temperatures (Table 3). This was the case for lentils (Lens culinaris), in which no effects were reported for dry heating at 120°C for 30 min, whilst autoclaving at similar temperatures (120°C for 20 min) increases the α-helices and random coil contributions. The latter treatment also reduces the contribution of the intramolecular β-sheets conformation in exchange for a larger amount of intermolecular β-sheets⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}. However, the response of the secondary structure of proteins to changes in moisture could be source-dependent, as not all the ingredients analysed follow the same pattern. Whilst a dry source of heat at 120°C for 30 min induces the appearance of random coil structures in kidney beans, the same thermal treatment does not induce the formation of these structures in lentils⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Carbonaro, Maselli and Dore ^{95 )}. These differences could be related to the proportion of the different types of proteins in each ingredient (for example, globulins v. albumins), or to the inherent structural characteristics of the proteins of each source. In these studies⁽ Reference Carbonaro, Maselli and Nucara ^{30 ,} Reference Carbonaro, Maselli and Dore ^{95 )} the seeds were autoclaved, along with the water that was used for soaking them (1:4, w/v) during 2 h.

Shear effects

Multiple definitions have been provided for shear. As defined for extrusion processing⁽ Reference Marsman, Gruppen and van Zuilichem ^{96 )}, shear is the dissipation of the mechanical energy input by friction of the particles inside the extruder barrel. The estimation of shear is based on the torque produced by the engine and on the calculation of specific mechanical energy (SME) input. Different levels of shear are involved in the equipment, which is commonly used for compound feed production. Pelleting consumes a low level of SME, ranging between 23 and 45 kJ/kg, and the levels are highly dependent on the fat level inclusion in the diet⁽ Reference Richardson and Day ^{97 ,} Reference Fahrenholz ^{98 )} and the size of the die. The technology that is considered to cause the highest levels of shear is extrusion, with SME ranging from 839 to 1277 kJ/kg⁽ Reference Fang, Zhang and Wei ^{68 )}. As a system parameter, torque is the result of the combination of several process parameters, such as temperature, moisture content, feeder-screw speed, extruder-screw speed and screw configuration. Thus, any change in the process parameters during extrusion can lead to an alteration of the SME input. Separating the effects of SME input from other process parameters used during processing in complex systems such as extrusion is difficult, and literature describing changes in protein structure due to shear only is scarce.

The effects of simple shear flow on protein structure and function have been reviewed before⁽ Reference Bekard, Asimakis and Bertolini ^{99 )}. However, most of these research has been performed on model systems of proteins in solution⁽ Reference Bekard and Dunstan ^{100 ,} Reference Dunstan, Hamilton-Brown and Asimakis ^{101 )}, which do not resemble the conditions used during feed production. Simple shear of proteins can induce conformational changes to the structure of proteins by disrupting the secondary/tertiary structure, even at relatively low temperature, as illustrated by the work of Dunstan and colleagues⁽ Reference Bekard and Dunstan ^{100 ,} Reference Dunstan, Hamilton-Brown and Asimakis ^{101 )}.

Protein aggregation in soya isolates has been reported as a consequence of extrusion⁽ Reference Fang, Zhang and Wei ^{68 )}. Vital factors during extrusion, such as a high temperature and pressure, were suggested to be responsible for protein aggregation⁽ Reference Fang, Zhang and Wei ^{68 )}. However, with increasing SME levels, the molecular weight of the protein aggregates was reduced. The molecular weight of the aggregates was reduced by 5, 9 and 17 % at SME inputs of 1050, 1093 and 1277 kJ/kg, respectively, compared with the molecular weight at the 839 kJ/kg extruder setting⁽ Reference Fang, Zhang and Wei ^{68 )}. It was suggested by Meade et al. ⁽ Reference Meade, Reid and Gerrard ^{54 )} that the mechanical forces during extrusion could break the peptide bonds in the amino acid chain (effects on primary protein structure), thus producing a similar effect as proteolysis.

Increased levels of non-covalent and covalent bonds, which lead to larger amounts of aggregation, were reported⁽ Reference Marsman, Gruppen and De Groot ^{102 )} in soya proteins with increasing SME. In this experiment, shear contrasts were produced by adaptations to the screw configuration of the extruder. With the screw configurations that produced increased levels of shear, there was also a decrease in the solubility after the extraction of proteins with urea or DTT and the successive extractions with solutions of urea/DTT or DTT/urea. This could indicate that after extrusion at high mechanical energy inputs, proteins may interact by mechanisms other than the formation of disulfide or non-covalent bonds. These mechanisms could be covalent bonds related to crosslinking between amino acids⁽ Reference Marsman, Gruppen and De Groot ^{102 )}.

The formation of Maillard reaction products (and possibly also protein structural modifications) is affected by the interaction between temperature and mechanical energy⁽ Reference Lei, Fulcher and Ruan ^{103 )}. High SME inputs produced large amounts of Maillard reaction products when extrusion of a rice (Oryza spp.)–glucose–lysine model mixture was performed at barrel temperatures of 100 and 130°C, but not at 70°C⁽ Reference Lei, Fulcher and Ruan ^{103 )}. Increasing the screw speed from 80 to 140 rpm had limited effect on the in vitro crude protein digestibility of commercial soyabean meal after extrusion⁽ Reference Marsman, Gruppen and Van der Poel ^{104 )}.

Although there is little evidence that crude protein digestibility might be largely affected after shear-processing, the formation of crosslinked or Maillard-modified amino acids has a large impact on the nutritional value of the feed. The crosslinked and modified amino acids cannot be utilised after absorption⁽ Reference Van Barneveld, Batterham and Norton ^{105 )}, and therefore animal performance is decreased.

Particle size reduction processing

The amount of mechanical energy involved in particle size reduction is related to the type of equipment employed. SME ranges from negligible in roller milling to 104·4–126 kJ/kg for hammer milling⁽ Reference Bitra, Womac and Chevanan ^{106 )} and 6·5–10·1 kJ/kg for the multicracker system⁽ Reference Thomas, Vrij and Zandstra ^{107 )}. Hammer and roller milling are the most frequently used technologies for particle size reduction of compound feed ingredients, although new technologies, such as the multicracker, have been developed⁽ Reference Thomas, Vrij and Zandstra ^{107 )}.

The effects of particle size reduction on digestibility has been analysed before⁽ Reference L’Anson, Choct and Brooks ^{21 ,} Reference Lahaye, Ganier and Thibault ^{108 –} Reference Valencia, Serrano and Lázaro ^{110 )}. During particle size reduction, rupture of protein bodies from grains and legume seeds that contain storage proteins increases protein solubility⁽ Reference Vishwanathan, Singh and Subramanian ^{111 )}, which in turn could facilitate enzyme accessibility. However, only few studies report improved enzyme accessibility due to a larger area of exposure of the substrate⁽ Reference Barakat, de Vries and Rouau ^{112 )}. Few studies have been performed on the effects of particle size reduction on protein structure. This is probably related to the fact that particle size reduction is regarded as only a physical change and, as such, it is assumed not to have major effects on the structural characteristics of proteins.

Some of the technologies that are used to reduce particle size of ingredients involve the use of moisture, whilst others are performed under dry conditions. Wet-milling of maize dried distillers grains with solubles increased in vitro crude protein digestibility⁽ Reference De Vries, Pustjens and Kabel ^{113 )}. It is possible, as suggested by these authors, that the disruption of the cell wall structure leads to increased enzyme accessibility and that the improved in vitro digestion is not linked directly to changes in protein structure, although the latter point was not measured by the authors. Supporting this suggestion, is the observation that there were no changes in the secondary protein structure after wet-milling of silk from Philosomia cynthia ricini and Bombyx mori silkworms⁽ Reference Rajkhowa, Wang and Kanwar ^{114 )}. In contrast, when dry milling was used in the same study, the intramolecular β-sheet structures in these proteins completely disappeared, probably due to the processes involved in dry milling, including shear. The silk in that study was wet (i.e. addition of water) or dry milled using a ball mill followed by air jet milling. In the case of wet milling, water could act as a lubricant, thereby decreasing the amount of shear. It is possible that the disappearance of the intramolecular β-sheets after dry milling could increase protein digestibility as previously reported by Carbonaro et al. ⁽ Reference Carbonaro, Maselli and Nucara ^{30 )}. Furthermore, dry milling could have the additional benefit of the disruption of cell wall structures (as mentioned before for wet-milling), which can increase the access of enzymes for protein hydrolysis.