Protein composition and physicochemical properties of soy protein isolates and their role in structure formation during high moisture extrusion

The ongoing (partial) transition from animal- to plant-based protein in the global food system has created a demand for high quality plant protein ingredients. Among plant protein sources, soybeans [Glycine max (L.) Merr.] are abundantly used due to their high protein content and nutritional quality. The protein transition has also impelled an increased interest in plant-based meat analogues that mimic the sensory properties of meat products. High moisture extrusion (HME) is a technique that allows obtaining products with a fibrous or layered structure from e.g. soy protein. However, structuring mechanisms occurring during HME processing are not fully understood. In systems with only one protein component, e.g. when soy protein isolate (SPI, protein content > 90%) is used, mainly alignment of protein strands and the occurrence of syneresis (i.e. water expulsion from a gel or protein network), both as a result of the formation of (non-)covalent interactions and bonds during the HME process, have been proposed to result in anisotropic structure formation. These hypotheses were mainly based on empirical studies conducted with commercial SPIs, which are typically characterized by high protein denaturation degrees, low dispersibilities and often have unknown (protein) compositions. Although these characteristics are frequently observed in commercial SPIs, information on the impact of the protein isolation conditions on the protein composition and physicochemical properties of SPI is lacking in literature. Furthermore, both the protein composition of SPI and its protein physicochemical properties have been shown to affect SPI protein network formation upon thermal treatment. However, their impact on structure formation during HME is largely unknown.

Against this background, the main objective of this doctoral dissertation was to increase overall understanding on the roles of soy protein isolate composition and physicochemical properties in determining structure formation during HME. To be able to reach this objective, in-depth understanding of factors determining SPI properties is imperative. Therefore, this dissertation also aimed to systematically investigate the impact of varying isolation conditions on the physicochemical properties and protein composition of SPI.

To be able to study the impact of processing on soy protein composition, first a size-exclusion chromatography (SE-HPLC) method was developed to quantitatively determine the soy protein composition at subunit level. Analysis under denaturing and reducing conditions allowed detection and quantification of proteins with molecular weight (MW) ≥ 97 kDa, 7S globulin α+α’- and β-subunits, 11S globulin A- and B-subunits and 2S albumins.

As proof-of-concept, this SE-HPLC method was first used to study variability in Belgian soy protein composition. Indeed, soybeans cultivated in Belgium were used as raw materials in the experimental work of this dissertation. To date, there is practically no information available on the impact of variety and growing conditions on the protein composition of Belgian soy. To this end, the composition of three soybean varieties (Sioux, Bettina and Lenka) as function of harvest year, growing location and inoculation strategy was studied. Both genotype and harvest year had a significant impact on the protein composition of Belgian soy samples, while the impact of growing location and the use of different inoculants or no inoculant was limited. All factors, however, had a significant impact on the proximate composition of soybeans. On average, Sioux samples had the highest protein content (51%) and the lowest 7S/11S globulin ratio (0.50), while Bettina samples had the lowest protein content (39%) and highest 7S/11S globulin ratio (0.65).

Sioux, Bettina and Lenka flour samples were then used to evaluate the impact of a conventional soy protein isolation procedure under varying isolation conditions, i.e. subsequently (i) hexane defatting, (ii) alkaline extraction (pH 7.0-9.0), (iii) acid precipitation (pH 3.5-5.5) and (iv) freeze/spray-drying, on the protein denaturation degree, dispersibility and composition of the obtained protein fractions. The latter was evaluated with the developed SE-HPLC method. Neither hexane defatting nor alkaline extraction induced protein denaturation. An increase in the alkaline extraction pH did, however, result in an increased protein extraction yield. Acid precipitation at pH 5.5 followed by neutralization resulted in SPIs with higher protein purities and dispersibilities, lower 7S/11S globulin ratios and 7S globulins with a slightly lower thermal stability compared to acid precipitation at pH 3.5 and 4.5. Notably, the observed pH-dependent trends were not influenced by the used soybean variety for the isolation procedure. All obtained SPIs had relatively high dispersibility (80-100%) and only (partial) 2S albumin denaturation was observed. Spray-drying did not induce protein denaturation or a reduction in protein dispersibility in comparison to freeze-drying. From these findings, it was concluded that the typically observed high degrees of protein denaturation and low dispersibilities of commercial SPIs can be ascribed to an additional heating step typically applied in industrial processes. Finally, it was notable that all SPI dispersions, regardless of the isolation conditions, contained a considerable amount of aggregated protein structures next to native globulins. This can impact the functional properties of SPI and structure-function relationships in this regard should be investigated further.

In the final part of this dissertation, the impacts of (i) SPI physicochemical properties and (ii) protein composition on the structure and texture of high moisture extrudates were assessed in two separate chapters.

First, to evaluate the impact of varying protein physicochemical properties, here understood as relating to differences in the degree of denaturation and the dispersibility of the proteins, in-house produced SPIs containing (i) native and dispersible proteins, (ii) denatured, non-dispersible and aggregated proteins, or (iii) blends thereof were used for lab-scale HME processing. Most native and dispersible SPI proteins took part in the formation of both non-covalent interactions and disulfide bonds during HME processing, which resulted in a layered extrudate microstructure with water-rich domains indicating the occurrence of syneresis. Even after disruption of non-covalent interactions and disulfide bonds, high MW aggregates (> 100 kDa) were detected in the extrudate extract using the earlier developed SE-HPLC method. The inclusion of denatured, non-dispersible and aggregated proteins besides native and dispersible proteins in the raw material resulted in higher extrudate hardness and cutting strength, but still resulted in a layered extrudate structure. HME processing of only denatured, non-dispersible and aggregated proteins resulted in an extrudate with a fibrous structure with no indication of syneresis and with similar textural properties as the extrudate based on native and dispersible protein. Furthermore, no pronounced further reduction in protein extractability or changes in MW distribution were observed upon extrusion of raw material containing denatured, non-dispersible and aggregated proteins. Therefore, it was hypothesized that high MW aggregates present in the raw material were not broken down pronouncedly during HME processing probably making them less available for forming new protein-protein interactions and bonds. This may also explain why no syneresis seemed to occur upon extrusion in this case. However, more research in this regard is necessary.

Second, to evaluate the impact of protein composition, fractions enriched in either soy 7S or 11S globulins were prepared and blended in different ratios for lab-scale HME processing. The fraction enriched in 11S globulins had liquid-like behavior when hydrated to 60% (w/w) moisture and its use for HME at the same moisture content resulted in extrudate with irregular morphology. Addition of the fraction enriched in 7S globulins, with a more elastic structure, to the raw material resulted in layered extrudate structures. All extrudates containing the fraction enriched in 7S globulins had similar textural properties. Extrudate based on the fraction enriched in 11S globulins was mainly stabilized by disulfide bonds, while with increasing 7S globulin content in the raw material the importance of non-covalent interactions increased. Disulfide bonds were mainly formed by 11S globulin A- and B-subunits. Some 7S globulins α/α’-subunits also participated in disulfide bond formation, but 7S globulins β-subunits were not involved. The extrudate high MW aggregates that were also detected in extrudate based on native and dispersible SPI proteins were found to be mainly composed of 11S globulin B-subunits if raw material enriched in 11S globulins was used and of 7S globulin α/α’-subunits if raw material enriched in 7S globulins was used.

In conclusion, the findings of this doctoral dissertation provided new insights in (Belgian) soy composition and factors affecting SPI properties, both of which are relevant for the functional properties of soy protein food ingredients. Furthermore, it can be concluded that mainly SPI physicochemical properties affected extrudate (micro)structure and texture. Differences in physicochemical properties and protein composition between commercial SPIs may well be responsible for the high variability observed in literature regarding the extent to which and the type of protein-protein interactions and bonds formed during HME processing.