Document [original]

This version is available at https://doi.org/10.14279/depositonce-10165

This work is licensed under a CC BY-NC-ND 4.0 License (Creative

Commons Attribution-NonCommercial-NoDerivatives 4.0 International). For

more information see https://creativecommons.org/licenses/by-nc-nd/4.0/.

Klost, M., Brzeski, C., & Drusch, S. (2020). Effect of protein aggregation on rheological properties of pea

protein gels. Food Hydrocolloids, 108, 106036. https://doi.org/10.1016/j.foodhyd.2020.106036

M. Klost, C. Brzeski, S. Drusch

Effect of protein aggregation on

rheolo

ical properties of pea protein

els

Accepted manuscript (Postprint)Journal article |

Effect of protein aggregation on rheological properties of pea

protein gels

M. Klost1, 2, C. Brzeski1 and S. Drusch1, 2

1 Technische Universität Berlin, Faculty III Process Sciences, Institute of Food technology and Food Chemistry, Department

of Food Technology and Food Material Science, Straße des 17. Juni 135, 10623 Berlin, Germany

2 NutriAct – Competence Cluster Nutrition Research Berlin-Potsdam

* Correspondence: [email protected]

ABSTRACT

Yoghurt style gels are a promising way to increase the consumption of plant derived proteins. However,

reaching texture properties similar to those commonly known from milk yoghurts while incorporating

large amounts of plant-derived proteins, is a challenge that needs to be addressed to meet consumers

expectations. Therefore, this study aims to investigate the influence of pH conditions (pH 6.0 to pH 8.0)

during pre-treatment on the rheological properties of fermentation induced pea protein gels with a

protein content of 10%. Results showed a strong correlation between the pH value during pre-treatment

and the protein solubility after pH readjustment to pH 8. Solubility was highest if pea protein was pre-

treated at pH 8.0 and lowest if it was pre-treated at pH 6.0. Since only soluble aggregates are believed

to participate in network formation, networks formed by pea protein pre-treated at pH 6 were coarser,

than those formed by pea protein pre-treated at pH 8 owing to a lower degree of crosslinking caused by

less available protein. Coarser networks and higher proportions of insoluble particles increased loss of

water and lowered the storage modulus G’ as well as the ability of the networks to recover after intense

shearing. In particular pre-treatment at pH values below 7.0 led to gels with storage moduli of the same

magnitude as those measured in commercial milk derived yoghurts. Adjusting the pH value during pre-

treatment of pea protein can therefore be considered a promising approach for the customisation of

texture properties while maintaining a constantly high protein content.

1. INTRODUCTION

In the light of increasing life expectancy, it is inevitable to address the issue of age related non-

communicable diseases in a preventive manner (WHO, 2013). One approach is to increase consumer

awareness for a healthy lifestyle and a balanced nutrition and to provide a range of corresponding foods.

In this context, plant derived proteins have been proposed to contribute to the prevention of chronic

degenerative diseases (Krajcovicova-Kudlackova, Babinska, & Valachovicova, 2005). As a

consequence, an adequate intake through consumption of plant -based foods needs to be achieved. Plant

protein-enriched beverages and emulsion products cannot deliver the required amounts of plant derived

protein for this purpose. In contrast, gels are dispersed systems more suitable for the incorporation of

relevant amounts of plant derived protein. E. g. yoghurt has a high consumer acceptance when it comes

to protein-rich foods (Banovic et al., 2018), and yoghurt type products from plant derived proteins allow

for the incorporation of up to at least 10% plant derived protein (Klost & Drusch, 2019; Klost, Giménez-

Ribes, & Drusch, 2020).

Plant-derived proteins (such as pea and soy) consist of hexameric 11S and trimeric 7S globular protein

fractions. The general process of acid induced gelation has been extensively described by a variety of

authors and for proteins from various plants. It starts with a heating step during which the protein

undergoes heat induced structural rearrangements that lead to the formation of aggregates. When

electrostatic repulsion is lowered during acidification, soluble aggregates form network structures by

hydrophobic interactions e. g. (Ringgenberg, Alexander, & Corredig, 2013). However, during heat

treatment – apart from the soluble aggregates required for gelation – insoluble aggregates may form.

Whether soluble or insoluble aggregates are formed depends on the sensitivity of the different protein

fractions to environmental factors like ionic strength, temperature and pH conditions (Yamagishi,

Miyakawa, Noda, & Yamauchi, 1983). In this context, changes in pH conditions or ionic strength may

decrease the electrostatic repulsion between two particles (Cano-Sarmiento et al., 2018) and

consequently promote short range particle interactions (Klemmer, Waldner, Stone, Low, & Nickerson,

2012). Generally, differences in sensitivity towards environmental factors can lead to a variety of

aggregates that are formed between different protein fractions and/or different protein fraction subunits

via disulphide or non-covalent bonds. Due to lack of detailed studies on the aggregation behaviour of

pea protein and vast similarities between pea and soy proteins we refer to literature on soy protein as an

indicator for pea protein’s aggregation behaviour. Generally speaking, various 7S and 11S fractions can

form aggregates with other 7S or 11S fractions via various types of interactions. A summary of soy

protein fractions and types of interactions involved in those soluble and insoluble aggregates is given in

table 1.

Table 1 types of aggregates formed upon heating of mixtures of soy 11S and 7S: types of interactions involved in aggregate

formation and composition of aggregates (corresponding pea proteins are: 11S

→

Legumin, 7Sαα’

→

Convicilin, 7Sβ

→

Vicilin)

subunits involved

interactions

reference

soluble

7S αα’

disulphide bond

(Yamagishi et al., 1983)

7S αα’ and 11S acidic

disulphide bond

(Yamagishi et al., 1983)

7S β and 11S basic

disulphide bond

(Damodaran & Kinsella, 1982; German,

Damodaran, & Kinsella, 1982)

7S αα’ and 7S β

no disulphide bond

(Yamagishi et al., 1983)

7S β

no disulphide bond

(Yamagishi et al., 1983)

7S β and 11S basic

no disulphide bond

(Petruccelli & Añón, 1995)

insoluble

11S acidic and basic

disulphide bond

(Yamagishi et al., 1983)

7S αα’ and 11S acidic+basic

disulphide bond

(Yamagishi et al., 1983)

7S β

no disulphide bond

(Yamagishi et al., 1983)

11S basic

no disulphide bond

e.g. (German et al., 1982)

Different ratios of soluble to insoluble aggregates may influence a protein’s ability to form fermentation

induced gels. While soluble aggregates are a prerequisite for the formation of fermentation induced gels

e. g. (Ringgenberg et al., 2013) insoluble aggregates may act as inactive fillers and weaken an emerging

gel matrix (Britten & Giroux, 2001). Weakened gel matrices should in turn be reflected in rheological

and texture parameters. In two previous studies (Klost & Drusch, 2019; Klost et al., 2020) with different

objectives – and therefore different environmental parameters such as pH, homogenisation pressure,

heating temperature and heating time during protein pre-treatment prior to fermentation – we found

relevant inter-study differences in rheological moduli G’ and G’’. While for the first study no specific

pre-treatment – apart from heating – was performed (Klost & Drusch, 2019), we additionally applied

enzymatic hydrolysis as a pre-treatment before fermentation in the second study (Klost et al., 2020). In

preliminary experiments from this study we tested different pH values (pH 7.0, 7.5 and 8.0) during

hydrolysis for one of the applied enzymes (Protamex®) to determine its pH optimum (unpublished data).

Interestingly, we found significant differences between the storage moduli G’ at the end of fermentation

but no differences in the molecular weight distribution of the corresponding hydrolysates

(supplementary fig S1). G’ decreased from 5052 ± 172 Pa (hydrolysis at pH 8.0) to 1487 ± 180 Pa

(hydrolysis at pH 7.0). Putting these values in line with the results from the first study (complex shear

modulus |G*| = 452 ± 27 Pa (corresponding to storage modulus G’ = 446 ± 26 Pa)) where heating was

conducted at pH ~6.5 we suspect the influence of pH value during heat pre-treatment to be the most

important parameter for these inter-study differences. However, this presumption needs to be confirmed

in a systematic investigation.

Therefore, the focus of this study is to investigate the application of pH variation during pre-treatment

of pea protein to specifically customise the rheological properties of subsequently produced yoghurt

alternatives. To this regard, we propose the following mechanism by which the pH value during pre-

treatment (heating and homogenising) influences the aggregation behaviour of pea protein: at a pH that

leads to reduced electrostatic repulsion close-range interactions lead to an increase in insoluble

aggregates. A higher proportion of insoluble aggregates will weaken the gel structure and will result in

less stable gels. With this in mind it should be possible to customise the rheological properties of

fermentation induced pea protein gels by targeted manipulation of environmental parameters during a

pre-treatment step prior to fermentation while maintaining the protein content constant at 10%.

Moreover, beyond the development of yoghurt alternatives, customising the texture of fermentation

induced pea-protein gels may lead to a variety of new products such as spreads, cream fillings for bakery

and confectionary products, etc. in the future.

2. MATERIALS AND METHODS

Fig 1 gives a general overview over

the experimental setup. In a first set of

experiments the difference in intrinsic

fluorescence before and after heating,

the ζ-potential and the protein

solubility, of the untreated protein

were analysed in dependence of the

pH value to gain a deeper

understanding of the unfolding

behaviour during heating, the

electrostatic properties and the

formation of insoluble aggregates

respectively. Slurries with a protein

content of 10% were prepared from

this raw material. In subsequent steps

the pH of the slurries was adjusted

according to the experimental setup,

the slurries were heated and

homogenised followed by pH

readjustment to pH 8. Afterwards

these pre-treated slurries were either

lyophilised for molecular weight

analysis and further protein solubility

tests, or fermented for subsequent

rheological, microscopic and loss of

water characterisation.

Materials

Pea protein concentrate (LOT-Nr.: 16041801) with a protein content of 78% was obtained from IGV

(Institut für Getreideverarbeitung) GmbH, Nuthetal, Germany. The lactic acid culture (YoFlex®; S.

thermophilus and L. bulgaricus) was kindly provided by Chr. Hansen, Hoersholm, Denmark. Gels and

buffer-solutions for SDS-PAGE analysis were purchased from BioRad Laboratories GmbH (München,

Germany). All other chemicals were of analytical grade and purchased from Merck and Sigma Aldrich

(Darmstadt, Germany) and Carl Roth GmbH + Co.KG (Karlsruhe, Germany).

Fig 1 flow chart of experimental setup

ζ-potential measurement of untreated pea protein

ζ-potential measurements were conducted to estimate the electrostatic repulsive properties of the

untreated protein in dependence of the pH-value. Measurements were carried out in triplicate in protein

solutions containing 0.3% (w/w) of the untreated protein prepared in 0.01 M phosphate buffer at pH 6.0,

6.5, 7.0, 7.5, 8.0 (Zetasizer Nano-ZS, Malvern Instruments GmbH, Herrenberg, Germany).

Determination of intrinsic fluorescence of untreated pea protein

Intrinsic fluorescence measurements were carried out in order to determine differences in the unfolding

behaviour of pea protein, when heated at different pH values. Protein solutions of 0.05% (w/w) protein

in 0.01 M phosphate buffer were prepared at pH 6.0, 6.5, 7.0, 7.5 and 8.0. Samples were measured in a

Cary Eclipse Fluorescence Spectrophotometer (Agilent Technologies, Victoria, Australia) at an

excitation wavelength of 290 nm and the emission wavelength was scanned between 300 and 400 nm.

Emission wavelengths were scanned before heating, followed by subsequent heating to 50 °C, holding

for 60 minutes, further heating to 80 °C, holding for 30 minutes and another scan of the emission

wavelength. All samples were prepared in triplicate. For evaluation the wavelengths at maximum

emission before and after heating were determined and the red shift during heating was calculated as the

difference between the two values.

Solubility of pea protein

Protein solubility was measured before and after the pre-treatment process and before fermentation.

More specific, the solubility of the raw material was measured at pH 6 to 8 with steps of 0.5, the

solubility of pre-treated and freeze-dried samples was measured at the pH value during pre-treatment as

well as at pH 8 representing the pH at the start of fermentation. To this purpose, suspensions with a

protein content of 5% were prepared and the pH values were readjusted to the required value with

0.1 M/1 M NaOH or 0.1 M/1 M HCl if necessary. Suspensions were then left to stir for 60 minutes. An

aliquot of the suspensions was used to determine the total protein content and another aliquot was

centrifuged at 10000xg for 15 minutes for the determination of the soluble protein fraction. Protein

contents were determined with a Dumatherm® (C. Gerhardt GmbH&Co. KG, Königswinter, Germany)

at an oxygen flow rate of 100 mL/min and 0.8 mg oxygen/mg sample. Protein solubility was then

calculated as

protein solubility=csoluble protein

ctotal protein ∙100% (1)

Pea protein pre-treatment, lyophilisation and fermentation

A protein slurry (10% protein (w/w)) was heated to 50 °C under constant stirring and the pH value was

adjusted to 6.0, 6.5, 7.0, 7.5 or 8.0 respectively with 1 M NaOH and/or 1 M HCl. After a holding time

of 60 minutes 3.75% (w/w) sugar was added to slurries for fermentation, the temperature was increased

to 80 °C and the sample was held for further 30 minutes followed by cooling to approximately 40 °C in

an ice bath. Samples for lyophilisation were prepared without the addition of sugar. The cooled slurries

were pre-homogenised (Ultraturrax T25 basic, IKA, Germany, 30 s, 17500 rpm), strained through a

sieve and high pressure homogenised (Panda Plus, Niro Soavi, Germany, 80 MPa, one run). To samples

for fermentation, starter culture (YC-X11 Yo-Flex®, Chr. Hansen, Hoersholm, Denmark) was added

subsequently and the samples were filled into centrifuge tubes for loss of water experiments, beaker for

confocal laser scan microscopy (CLSM) and disposable rheology cups (Cat No 3716, Anton Paar, Graz,

Austria) for rheological tests. Samples were then fermented in a water-bath at 43 °C for 18 h. After

fermentation, the samples were stored at 4 °C for 24 hours before further investigation. Samples for

lyophilisation were frozen in an ethanol bath after homogenisation followed by lyophilisation. All

samples were prepared in triplicate for fermentation and lyophilisation. Additionally, one repetition of

each sample was prepared for CLSM experiments.

Molecular weight distribution of pre-treated and lyophilised pea protein via SDS-PAGE and size

exclusion chromatography

SDS-PAGE on 12% CriterionTM TGXTM Gels (26 wells) (BioRad Laboratories GmbH, München,

Germany) was used to characterise the molecular weight profiles of pre-treated and lyophilised samples.

Experiments were conducted according to the BioRad Bulletin #4110001 under reducing and non-

reducing conditions with Biorad 10xTris/Glycine/SDS (Cat# 161-0732) as running buffer. Sample

concentration was 0.1% in sample buffer (Biorad 2xLaemmli sample buffer, Cat# 161-0737 with or

without addition of dithiothreitol) and 10 µL of the samples were applied to the gels alongside a

molecular weight marker (PageRuler™ Prestained Protein Ladder, Cat# 26616, ThermoScientific).

Evaluation of the gels was carried out via photography of the gels followed by band identification via

estimation of their position in relation to the marker in combination with reference values from literature

and transformation to peaks for presentation (open source software ImageJ 1.52d (Schneider, Rasband,

& Eliceiri, 2012)). Transformation to peaks was done as mean of three gel lanes.

Size exclusion chromatography (SEC) of lyophilised samples (0.1% (w/w) in 0.1 M phosphate buffer,

pH 8) was performed in triplicate on a Superdex 200 Increase 10/300 GL (GE healthcare GmbH,

Solingen, Germany) column with 0.1 M phosphate buffer as eluent (HPLC ÄKTAbasicTM 10 system,

Amersham Biosciences, Uppsala, Sweden). Detection took place via an UV detector at 280 nm.

Qualification of peaks was not possible, but determination of the calibration area was performed by

using the highest and lowest calibration points from previous experiments.

Confocal laser scan microscopy (CLSM) of fermentation induced pea protein gels

For CLSM, 20 µl rhodamine B solution (0.2% (w/w) in distilled water) per gram sample were added to

the protein suspension before fermentation. CLSM was performed on one set of fermented samples. The

microscope was a Leica SP8 (Leica Microsystems GmbH, Wetzlar, Germany) with a HC PL APO CS2

63x/1.20 water objective (pinhole at airy unit 1 AU (111.5 µm)). Image resolution was 512x512 pixels.

For GFP detection a 3% laser (552 nm) intensity was coupled with emission detection of 580 nm at a

gain of 357. The number of required z-stacks was determined using the system optimised calculation of

z-stacks.

Loss of water

The fermented samples in the centrifuge tubes were centrifuged (500 g, 20 °C, 10 minutes, Avanti J-E,

Beckman Coulter GmbH, Krefeld, Germany) in a method adapted from (Guzmán-González, Morais,

Ramos, & Amigo, 1999). Subsequently the supernatant was discarded and the remaining sample was

weight. Loss of water was calculated as:

loss of water = masstotal-masspellet

masstotal ∙100% (2)

Rheology

Determination of all rheological properties was carried out in triplicate on an MCR 502 (Anton Paar,

Austria, concentric cylinder system CC 27 (measuring bob radius = 13.33 mm, measuring cup

radius = 14.46 mm, gap length = 40 mm)). For time-sweeps additional rheometers were used (Physica

UDS and MCR 301, Anton Paar, Austria, concentric cylinder system Z 3 DIN (measuring bob

radius = 12.5 mm, measuring cup radius = 13.56 mm, gap length = 37.5 mm) and CC 27 (measuring

bob radius = 13.33 mm, measuring cup radius = 14.46 mm, gap length = 40 mm) respectively). Special

care was taken, that replications of each sample were performed on at least two different rheometers.

Time sweeps were carried out during fermentation at 43 °C for 18 hours (f = 1 Hz, γ = 0.1%) in order

to track the structuring process. G’ and tan δ were chosen as parameters for evaluation. Thixotropy tests

were performed according to DIN SPEC 91143-2, 2012: samples were oscillated (f = 1 Hz, γ = 0.1%)

for 120 s followed by shearing for 120 s at γ󰇗 = 200 s-1 and oscillation for another 300 seconds. From

these experiments recovery of structure was calculated as:

recovery= G'End

GStart

'∙100%. (3)

Frequency sweeps were conducted at γ = 0.1% and frequencies ranging from 10 Hz to 0,01 Hz. For

evaluation the slopes dlogG’ / dlogω from the double logarithmic plots were determined and compared.

For the characterisation of non-linear deformation behaviour, amplitude-sweeps were performed at

f = 1 Hz and strain amplitude γ0 between 0.01% and 1000%. First of all, the end of the linear viscoelastic

regime was determined as the point, where G’ varies more than 5% from the original value. Evaluation

of the non-linear deformation was carried out via Lissajous plots, stress decomposition and calculation

of the stiffening ratio (S-factor) and dissipation ratio φ. In this context, information about both inter-

and intracycle behaviour can be derived from Lissajous plots and the interpretation of the elastic stress

curves can contribute to understanding changes within the network structure by application of a model

that links rupture of colloidal gels to the bond number between individual particles in the gel network

e.g. (Hsiao, Newman, Glotzer, & Solomon, 2012; Park & Ahn, 2013; Park et al., 2015) or the description

of microcracks that occur prior to the complete rupture of gels (Faber, Van Breemen, & McKinley,

2017). Consequently, perfectly elastic behaviour – represented by a straight line in the intracycle

strain γ-elastic stress τ’ diagram – is related to a rigid cluster structure with high bond numbers (i.e. 4 to

6) in the gels (Park et al., 2015). At intercycle strain amplitude γ0 above the linear viscoelastic regime,

shear and strain amplitude begin to interfere with the network structures. Depending on the applied

model, this may be reflected in the decrease of bonding numbers e.g. (Hsiao et al., 2012; Park & Ahn,

2013; Park et al., 2015) or the occurrence of microcracks (Faber et al., 2017). Both would reduce the

size and volume fraction of rigid clusters which in turn leads to a reduction in the load bearing network

that would be capable of supporting elastic stress (Hsiao et al., 2012). If such behaviour occurs, it is

reflected in the onset of deviation of the curves from a straight line. This deviation often leads to an

inversed sigmoidal shape. This shape can be interpreted as follows: the decline of the slope at small

intracycle strain γ indicates a decrease in the ability to support elastic stress owing to the reduced size

and volume fraction of rigid clusters (Park et al., 2015) and therefore relates to overall intercycle strain

softening with increasing intercycle strain amplitude γ0. The increase of the slope at higher intracycle

strain γ can be related to the stretching of remaining rigid clusters which in turn causes strong intracycle

elasticity (Park et al., 2015) and indicates intracycle stiffening. The calculation of S-factors reflects this

intracycle behaviour onto the entire range of intercycle strain amplitudes γ0. S-factors were calculated

according to (Ewoldt, Hosoi, & McKinley, 2008):

S≡G'L-G'M

G'L (4)

where G’L is the large strain modulus (secant line of elastic Lissajous plot) and G’M is the minimum

strain modulus (slope of elastic Lissajous plots at zero). In this context an S-factor S > 0 indicates

intracycle strain stiffening, whereas S < 0 refers to intracycle strain softening (Ewoldt et al., 2008).

However, when evaluating S-factors the special case of pseudoplastic and elastoviscoplastic materials

needs to be considered. While in truly strain stiffening systems the overall stress will increase towards

higher intracycle strains γ (elastic Lissajous plot) as shown by Park et al in their fig 11 (Park, Ahn, &

Lee, 2015) and by Ewoldt et al in their fig 10 (Ewoldt, Winter, Maxey, & McKinley, 2010) and the S-

factor may become S > 1, in pseudoplastic and elastoviscoplastic materials the overall stress may

approach a perfectly rectangular shape in elastic Lissajous plots (Ewoldt et al., 2010) and the S-factor

will trend towards a maximum value of one. Especially in the latter case the elastic stress curve will be

horizontal at low intracycle strain γ and its slope may increase towards higher intracycle strain γ. This

leads to the S-factors trending towards one – and therefore S > 0 – despite a strain softening overall

rheology (Mermet-Guyennet et al., 2015). This effect is owing to the mathematical definitions of the S-

factor (Ewoldt et al., 2010). More specifically, in this case the tangent modulus at minimum intracycle

strain γ (GM) approaches zero and therefore equation 2 reduces to

S-factor = GL

GL = 1 (Ewoldt et al., 2010) . (4a)

In order to distinguish between true intracycle strain stiffening and a shift from predominantly elastic to

mainly plastic behaviour it is therefore important, to consider the S-factor in combination with the

dissipation ratio φ (Ewoldt et al., 2010)

φ= ED

ED,pp =πγ0G''

4τmax (5)

where ED is the dissipated energy per cycle and corresponds to the area enclosed by the elastic Lissaous

plot, ED,pp is the dissipated energy in the corresponding perfect plastic system, γ0 is the intercycle strain

amplitude, G’’ is the loss modulus at that stain amplitude and τmax is the maximum shear stress in the

considered oscillatory cycle. In this context the dissipation ratio φ relates the dissipated energy in the

sample to the dissipated energy in a corresponding perfectly plastic material and consequently allows to

categorise rheological behaviour into elastic (φ → 0) or plastic (φ → 1) behaviour with a known critical

value (φ = π/4) for Newtonian behaviour (Ewoldt et al., 2010).

All data was obtained and – except for the dissipation ratio φ – automatically calculated by

RheoCompass™ Software (Anton Paar, Austria)

3. RESULTS & DISCUSSION

Protein characterisation

Electrostatic interactions and unfolding properties of untreated pea protein

The ζ-potential reflects electrostatic interactions between individual protein molecules and depends on

the pH value of the surrounding medium. Results show, that the absolute value of the ζ-potential

significantly decreased from |20.4| ± 0.5 mV to |11.7| ± 1.0 mV with decreasing pH values during pre-

treatment (table 2) indicating lower electrostatic repulsion at lower pH values. However, it is generally

accepted, that ζ-potentials above |30| mV are a prerequisite for full electrostatic stabilisation. As a

general rule, stability of hydrocolloid stabilised oil-droplets with ζ-potentials below |15| mV cannot

exclusively be explained by double-layer repulsion (Dickinson, 2009) and (Piorkowski & McClements,

2014) recommend ζ-potentials above |20| mV for long-term stability of electrostatically stabilised

beverage emulsions.

Table 2 ζ-potential and protein solubility before heating, protein solubility of heated, homogenised and lyophilised protein at

the original pH value and at pH 8 simulating the beginning of fermentation and red shift during heating of the protein.

sample

ζ-potential before

heating [mV]

Protein solubility

unheated samples [%]

Protein solubility

heated samples [%]

Protein solubility

heated samples at pH 8 [%]

red shift during

heating [nm]

pH 6.0

-11.7a

1.0

20.8a,1

0.4

17.5a,2

0.9

37.6a,3

1.1

10.0

(♦)

pH 6.5

-14.2b

0.6

26.0b,1

1.5

21.1a,b,2

0.3

40.0a,b,3

2.1

7.7a

4.0

pH 7.0

-17.1c

0.8

35.2c,1

1.9

27.4b,2

1.4

44.7a,b,3

3.1

8.3a

1.2

pH 7.5

-19.9d

0.8

63.9d,1

0.1

40.3c,2

3.4

47.1b,2

3.2

11.0a

2.0

pH 8.0

-20.4d

0.5

68.8e,1

1.2

57.2d,2

4.0

57.2c,2

4.0

7.0a

3.6

Different letters represent significant differences (α=0.05) within columns, different superscript numbers represent significant differences

(α=0.05) within rows as derived from ANOVA followed by Tukey post-Hoc test.

(♦) value calculated from double determination due to equipment failure. Data point was excluded from ANOVA and post-Hoc test.

We therefore need to assume, that electrostatic stabilisation is insufficient to prevent flocculation at pH

6.0 and 6.5 and is inadequate to fully stabilise a protein dispersion at pH 7.0 and 7.5. In turn we can

assume, that the chosen pH-range is suitable to produce samples with differently pronounced

electrostatic repulsion, which may in turn lead to differences in aggregation. While the degree of

electrostatic repulsion determines how close individual particles may get to each other, the aggregation

itself may take place via further non-covalent interactions and via disulphide-bonds. The ability to form

these types of interactions and bonds in turn strongly depends on the accessibility of relevant protein

side chains and therefore on the protein unfolding. Protein unfolding can be determined via intrinsic

fluorescence measurements. Generally, a red shift i.e. an increase in the wavelength at emission

maximum represents the unfolding of a protein as indicated by the decrease in the interactions of

tryptophan residues with quenching groups and thus an increase in its exposure to the solvent (Cairoli,

Iametti, & Bonomi, 1994). In our study the red shift by 7 to 11 nm upon heating of protein solutions

from 20 °C to 80 °C (table 2) was not significantly dependent on the pH value during pre-treatment. We

therefore propose similar unfolding kinetics during heating in all samples. Consequently, the

electrostatic repulsion – or lack thereof – must be the main influence factor on aggregation behaviour.

Protein solubility of untreated, pre-treated and pH readjusted pea protein

Solubility experiments were carried out at three conditions. First of all, solubility of the untreated

samples was measured after pH adjustment to pH 6.0, 6.5, 7.0, 7.5 and 8.0 to characterise the individual

influence of pH. In a subsequent step the pre-treated, lyophilised samples were re-dispersed at the pH

value of their respective pre-treatment to characterise the additional impact of heating and

homogenisation under those pH conditions. Last but not least solubility of the lyophilised samples re-

dispersed at pH 8.0 was measured to differentiate irreversible loss of solubility from reversible loss and

to simulate conditions at the beginning of fermentation.

Solubility of the untreated protein showed a significant pH dependency (table 2). Solubility decreased

with decreasing pH and correlated with the ζ-potential (R=-0,939, table 3). Similar behaviour has been

extensively described for various plant derived proteins e.g. (Barac et al., 2010) and can be ascribed to

the formation of insoluble protein aggregates due to decrease in electrostatic repulsion with decreasing

pH value. In a second step, the solubility of pre-treated, lyophilised samples at the pH-value of their

respective pre-treatment was measured. Compared to the untreated protein at respective pH values, a

further significant decrease of protein solubility (table 2) was found, indicating an additional influence

of the pre-treatment process on protein solubility. Similar behaviour was previously reported for heating

of soybeans (Nishinari, Fang, Guo, & Phillips, 2014).

Table 3: correlation matrix for results that showed significant differences in tables 2&4

pH at

heating [-]

Solubility

(unheated)

[%]

Solubility

(heated,pH 8)

[%]

G‘end

[Pa]

G‘24h

[Pa]

recovery

[%]

loss of

water [%]

ζ-potential

[mV]

pH at heating [-]

1.000

solubilityunheated [%]

0.959

***

1.000

solubilitypH 8 [%]

0.885

***

0.842

***

1.000

G‘end [Pa]

0.962

***

0.971

***

0.892

***

1.000

G‘24h [Pa]

0.932

***

0.942

***

0.910

***

0.981

***

1.000

recovery [%]

0.911

***

0.873

***

0.771

***

0.871

***

0.860

***

1.000

loss of water [%]

-0.718

***

-0.602

-0.576

-0.581

-0.564

-0.704

***

1.000

ζ-potential [mV]

-0.966

***

-0.939

***

-0.834

***

-0,913

***

-0,891

***

-0,925

***

0,746

***

1.000

1rst Peak area (SDS)

0.891

***

0.851

***

0.703

***

0.797

***

0.730

***

0.852

***

-0.582

-0.884

***

* α=0.1 ** α=0.05 *** α=0.01

Finally, from solubility experiments with lyophilised samples readjusted to simulate the starting pH of

fermentation (pH 8) we obtained the following information. First of all, increasing the pH value

increased the solubility compared to the values measured at the pH of pre-treatment and secondly, we

found a decrease of solubility compared to the untreated sample at pH 8. The former can be ascribed to

the presence of some pH reversible aggregates, that dissolve owing to the increased electrostatic

repulsion upon increasing the pH. At the same time, the latter indicates an increase of insoluble

aggregates caused by irreversible protein denaturation during the pre-treatment. This increase of

insoluble aggregates was most pronounced in samples pre-treated at pH 6.0 and led to overall solubilities

between 37.6 ±1.1% (pre-treatment at pH 6.0) and 57.2 ± 4.0% (pre-treatment at pH 8.0, table 2) with a

correlation coefficient of R=0.885 (table 3). This indicates a shift in the proportions of soluble and

insoluble fractions caused by the pH-value during pre-treatment. Higher protein solubility can be

ascribed to a higher number of soluble aggregates and vice versa. More specific this means, that samples

pre-treated at lower pH values contain fewer soluble and more insoluble aggregates than samples pre-

treated at higher pH values. Since soluble aggregates are a prerequisite for gelation while insoluble

aggregates may act as inactive fillers (Britten & Giroux, 2001), the differently pre-treated samples are

expected to exhibit differences in gelation behaviour.

Molecular weight distribution

SDS-PAGE is most suitable to investigate individual protein sub-fractions involved in aggregation since

samples lose their quaternary structure and any non-covalent protein-protein interactions during sample

preparation. Moreover, if SDS-PAGE is performed under reducing and non-reducing conditions,

insights in the presence and constitution of disulphide bound aggregates can be gained. However, SDS-

PAGE does not distinguish between soluble and insoluble aggregates. To this purpose SEC can be used

to investigate the undenatured protein molecules and the formation of soluble aggregates. Consequently,

applying both methods leads to a more detailed understanding of the aggregation behaviour of pea

protein upon pre-treatment under different pH conditions and the protein fractions involved. For both

types of investigation, the lyophilised samples were re-dispersed at the starting pH of fermentation

(pH 8.0).

SDS-PAGE under reducing and non-reducing

conditions (fig 2a) shows all bands typically

expected in pea protein. In more detail, the major

pea protein fractions are convicilin at ~70 kDa

(Crévieu et al., 1997; Croy, Gatehouse, Tyler, &

Boulter, 1980; Swanson, 1990), legumin at

~60 kDa (Croy, Gatehouse, Evans, & Boulter,

1980) and vicilin at ~50 kDa (Gatehouse, Croy,

Morton, Tyler, & Boulter, 1981; Gatehouse,

Lycett, Croy, & Boulter, 1982; Gatehouse, Lycett,

Delauney, Croy, & Boulter, 1983). Moreover,

legumin consists of an acidic α-chain (MW~38-

40 kDa) and a basic β-chain (MW~20 kDa) (Croy,

Derbyshire, Krishna, & Boulter, 1979; Croy,

Gatehouse, Evans, et al., 1980), that are connected

via a disulphide-bond and appear as separate bands

on SDS-PAGE under reducing conditions. Vicilin

on the other hand is prone to posttranslational

autolysis which leads to various subunits of lower

molecular weights (Dziuba, Szerszunowicz, Nalecz, & Dziuba, 2014; Gatehouse et al., 1982).

Generally, fig 2a shows similar molecular weight profiles for all samples, independent of pH value

during pre-treatment. Besides the bands regularly associated with pea protein, all samples contain two

Fig 2 SDS-PAGE under reducing and non-reducing conditions

(a) and SEC (b) of pea protein heated at pH 6 to 8.

fractions of protein > 170 kDa. Out of these, the fraction with lower molecular weight appears to remain

unaffected by the pH value during pre-treatment and shows no relevant differences between reducing

and non-reducing conditions. However, upon repetition of the experiment for a random sample with

increased SDS-content in the sample buffer (results not shown) this fraction was decreased under

reducing conditions, indicating aggregates held together by a mixture of disulphide bonds and strong,

non-covalent interactions. The fraction with higher molecular weight only occurs as a prominent peak

under non-reducing conditions and corresponds to fractions of the sample that did not migrate into the

gel at these conditions. Since this peak is not present under reducing conditions, the previously retained

aggregates must have been formed via disulphide bonds. Disulphide bonds can stabilise both, soluble

and insoluble aggregates (table 1). Taking the results from solubility experiments (higher solubility at

higher pH during pre-treatment) into account, the larger amount of retained fraction at higher pH during

pre-treatment indicates an increased contribution of soluble disulphide bound aggregates.

Considering the aggregation behaviour known for soy where the 7S αα’ (corresponding to pea

convicilin) fraction forms soluble, disulphide bound aggregates with itself or the acidic 11S

(corresponding to the legumin-α subunit) fractions (Yamagishi et al., 1983) (table 1) these soluble

aggregates may be constituted from the corresponding pea protein fractions convicilin and legumin α.

However, if legumin α subunits were involved, leftover legumin-β subunits (~20 kDa) should appear

under non-reducing conditions. As this is not the case, we propose soluble disulphide bound aggregates

to only consist of the convicilin fraction. Moreover, given the small differences in convicilin peak

heights under reducing and non-reducing conditions, this type of aggregates is unlikely to be exclusively

responsible for the retained, disulphide bound fractions. We therefore propose a mixture of soluble

disulphide bound convicilin aggregates and various insoluble disulphide bound aggregates consisting of

legumin and convicilin (Yamagishi et al., 1983) (table 1).

In SEC (fig 2b, blue line), the untreated pea protein included various fractions within the calibration

area while results from pre-treated samples only showed very small fractions and fractions larger than

440 kDa. This indicates heat induced aggregation of legumin (~360 kDa (Croy et al., 1979)) vicilin

(~150 kDa (Gatehouse et al., 1981)) and convicilin (280 kDa (Croy, Gatehouse, Tyler, et al., 1980)).

Owing to the preliminary filtration step in the method, aggregates detected in SEC can be considered

soluble and are therefore likely to consist of various combinations of vicilin, convicilin and convicilin

plus legumin β (table 1). Fig 2b shows a decrease in these soluble aggregates with decreasing pH value

during pre-treatment as indicated by the decrease in peak size of the peak at ~8 minutes, again supporting

the conclusions drawn from protein solubility experiments and SDS-PAGE.

In summary, the combined results from protein solubility experiments, SDS-PAGE and SEC show a

decrease in the number of disulphide-stabilised and soluble aggregates with decreasing pH during pre-

treatment. Insoluble aggregates that may be stabilised via other types of interactions (especially

hydrophobic ones) cannot be determined with either method due to sample preparation.

Gel characterisation

Kinetics of rheological parameters and pH value during fermentation

During fermentation, the pH value dropped from pH 8 to pH values around 4.8. Pre-treatment at

different pH values did not significantly influence the kinetics or final pH value (table 4, fig S2a

supplementary). In contrast, storage moduli (G’) at the end of fermentation increased significantly with

increasing pH values during pre-treatment (table 4 and fig S2b supplementary) and show a strong

correlation (R = 0.842, table 3) to the protein solubility at the start of fermentation. Higher pH values

during pre-treatment led to an increased number of soluble aggregates which are a prerequisite for

network formation, while lower pH values led to an increased number of insoluble aggregates that may

act as inactive fillers (Britten & Giroux, 2001). Larger numbers of soluble aggregates will lead to a

higher degree of crosslinking and therefore to the formation of denser network structures with higher

storage and loss moduli. In contrast, a larger proportion of insoluble aggregates reduces the number of

soluble aggregates, which leads to a lower degree of crosslinking. Insoluble aggregates may additionally

disturb the network formation by acting as inactive fillers. In combination, this leads to coarser network

structures with lower moduli.

Table 4 pH, storage modulus G’ and loss factor tan δ at the end of fermentation, G’, recovery from thixotropy test, slope

dlogG’/dlogω from frequency sweeps and loss of water after 24 hours of gel storage.

sample

pHend [-]

G‘end [Pa]

tan δend

G‘24h [Pa]

recovery [%]

dlogG‘/dlogω

loss of water

[%]

pH 6.0

4.75a

0.10

164a

0.167a

0.006

517a

164

14.1a

2.0

0.11a

0.01

4.8a

2.0

pH 6.5

4.77a

0.03

281a

0.158a

0.015

459a

21.3

(♦)

0.14a

0.03

2.1(♦)

pH 7.0

4.79a

0.08

1623b

0.164a

0.005

2275a

238

28.7b

3.7

0.11a

0.02

0.9b

0.3

pH 7.5

4.82a

0.05

3665c

320

0.147a

0.006

4712b

678

32.0b

0.7

0.12a

0.00

1.0b

0.3

pH 8.0

4.77a

0.02

5488d

325

0.153a

0.003

7285c

1565

33.0b

2.6

0.11a

0.01

0.9b

0.8

Different letters represent significant differences (α=0.05) as derived from ANOVA followed by Tukey post-Hoc test.

(♦) value calculated from double determination due to equipment failure. Data point was excluded from ANOVA and post-Hoc test.

The differences in absolute values of G’ and G’’ caused by different pre-treatments did not affect the

ratio between elastic and viscous proportions. Loss factor tan δ was similar for all samples (table 4),

indicating similar viscoelastic network-properties under linear viscoelastic conditions at all pre-

treatment conditions. If the viscoelastic network properties of gels produced from the same raw material

under the same gelation conditions are similar while the absolute value of storage and loss modulus G’

and G’’ varies, it is reasonable to assume a similar general gelation mechanism based on similar types

of interactions, where the main difference between samples is the number of available soluble aggregates

and resulting degrees of crosslinking.

Characterisation of pea protein gels after resting at 4 °C for 24 hours

This proposed gel structure with different degrees of crosslinking was further investigated by small and

large amplitude rheology, loss of water and CLSM experiments.

Frequency sweeps are generally used as an indicator for time dependent deformation of samples. In this

context, short time behaviour is reflected by higher frequencies, and long-term behaviour by lower

frequencies. In our study – independent of pre-treatment conditions – all samples showed a similar, low

slope of log G’ over log ω (table 4 and fig S3

supplementary) for the investigated frequency

range. This is typical for gels and dispersions

(Mezger, 2006). Furthermore, the values of ~0.12

were of the same magnitude as in our previous work

(Klost & Drusch, 2019; Klost et al., 2020) and are

in line with values reported for milk yoghurts

(Hassan, Ipsen, Janzen, & Qvist, 2003). At lower

frequencies, the storage modulus additionally

indicates the degree of crosslinking. The higher the

modulus, the higher the degree of crosslinking and

vice versa (Mezger, 2006). For our samples this

implies an increase in the degree of crosslinking and

network density with increasing pH value during

pre-treatment. Moreover, differences in network

density were also found in CLSM experiments

(fig 3). In gels made from protein pre-treated at

pH 6.0 mainly large protein fragments are apparent

with only a coarse network-structure visible. With

increasing pH values during protein pre-treatment,

the corresponding gels become denser and the

number of large particles decreases.

From thixotropy experiments we derive differences in the restructuring ability of the gels (table 4). The

higher the pH during pre-treatment, the more pronounced was the structure recovery. Values ranged

from 14.1 ± 2.0% at pH 6.0 to 33.0 ± 2.6% at pH 8.0 with pH 7.0 to pH 8.0 showing significantly higher

recovery than samples pre-treated at pH 6.0. We ascribe these differences in structure recovery to the

lower degree of crosslinking in samples pre-treated at lower pH values. In gels with lower degrees of

crosslinking, the remaining network fragments are less likely to find a suitable counterpart for

restructuring after intense shearing. Owing to their inactive filler properties, insoluble aggregates may

additionally enhance this effect. In combination this leads to a decreased structure recovery. The results

from thixotropy experiments correlate (R = -0.704, table 3) with the results from loss of water

Fig 3 CLSM micrographs of fermentation induced pea

protein gels made from pea protein slurry pre-treated at

pH 6.0 to 8.0. pH values during pre-treatment are noted in

the upper left-hand corner of the micrographs.

experiments. Moreover, in loss of water experiments, samples made from protein pre-treated at pH 6.0

significantly differed from samples pre-treated at pH values ≥7.0. This increased loss of water after

protein pre-treatment at low pH values is caused by larger pores in the coarser network structure of the

corresponding gels.

Amplitude sweeps (fig 4) were performed to characterise the non-viscoelastic behaviour of the gels.

Outside the linear viscoelastic regime, the rheological behaviour cannot exclusively be described by G’

and G’’ (fig 4a, top row) as higher harmonics become more relevant (Hyun et al., 2011). Appropriate

additional means to interpret the transition from linear viscoelastic to non-linear viscoelastic behaviour

as well as the non-linear viscoelastic behaviour itself are Lissajous plots (fig 4b), stress decomposition

(fig 4c), dissipation ratio φ (fig 4a, middle row) and the calculation of the stiffening ratio (S-factor)

(fig 4a, bottom row). Especially with regard to dissipation ratio φ and S-factor it needs to be kept in

mind, that they describe the intracycle behaviour at a fixed intercycle strain amplitude γ0, rather than the

overall intercycle behaviour. Lissajous plots can be used to interpret the overall intercycle behaviour of

gels as well as the intracycle deviation from linear viscoelastic behaviour by analysing their rotational

behaviour and overall shapes. From stress decomposition a closer insight into the changes to elastic

stress and more detailed knowledge on the intracycle stiffening/softening behaviour can be obtained.

The calculation of S-factors reflects this intracycle behaviour onto the entire range of intercycle strain

amplitudes γ0 and the dissipation ratio φ can be applied to distinguish intracycle strain stiffening

behaviour from effects caused by the transition from predominantly elastic to mainly plastic behaviour.

Results for all investigated samples in our study are shown in fig 4. Fig 4a shows the intercycle

development of storage modulus G’ and loss modulus G’’ (top row), alongside the projection of

intracycle parameters dissipation ratio φ and stiffening ratio (S-factor) on the entire range of investigated

intercycle strain amplitudes γ0 (middle and bottom row respectively). Fig 4b uses Lissajous plots to

show the total stress τ over the intracycle strain γ for all samples at various intercycle strain

amplitudes γ0 and fig 4c shows the corresponding elastic stress τ’ over the intracycle strain γ for all

samples at various intercycle strain amplitudes γ0.

Analogous to observations in all other rheological tests, results from amplitude sweeps showed similar

curves of storage modulus G’ and loss modulus G’’ for all samples (fig 4a, top row). Between samples,

these curves mainly differed in the absolute values of G’ and G’’. Linear viscoelastic regimes –

calculated as intercycle strain amplitude γ0 up to which G’ deviated no more than 5% from the value at

the lowest intercycle strain amplitude γ0 – extended to γ0 ≈ 1% with no relevant influence of pH during

pre-treatment.

Fig 4 amplitude sweeps

of fermentation induced

pea protein gels made

from pea protein pre-

treated at pH 6.0 to 8.0

at intercycle strain

amplitudes γ0 between

0.1% and 1010% and a

frequency of 1 s-1. (a)

top row: storage and

loss modulus G’ and G’’

over, γ0, (a) middle row:

dissipation ratio φ over

γ0, (a) bottom row:

stiffening ratio (S-

factor) over γ0, (b):

elastic Lissajous plots,

The extend of the linear viscoelastic regime is further supported by results from the dissipation ratio φ

and S-factor, where constant values are remained up to intercycle strain amplitudes γ0 between 1% and

2.5% (fig 4a, second and third row). Beyond the viscoelastic regime the overall intercycle rheological

behaviour is strain softening. However, in order to take higher harmonics into account and to determine

potential differences between samples it is worthwhile to also investigate the intracycle rheological

behaviour by means of the parameters illustrated above.

At intercycle strain amplitudes γ0 within the linear-viscoelastic regime (γ0 ≤ 1%, marked “I” in fig 4a

to c) elastic Lissajous plots (fig 4b I) have distinct elliptical shapes and the elastic stress τ’ assumes the

shape of a straight line (fig 4c I) as expected for linear viscoelastic behaviour. Moreover, no differences

were observed between samples and the narrow shape of the Lissajous plots indicates predominantly

elastic properties. This corresponds to the tan δ values, which are closer to zero than to one (table 4) and

to dissipation ratios φ of approximately 0.2 (fig 4a I, middle row) that indicate predominantly elastic

behaviour.

Lissajous plots and elastic stress τ’ at intercycle strain amplitudes 1% ≤ γ0 ≤ 6.3% (fig 4b and c II), begin

to rotate clockwise with increasing intercycle strain amplitude γ0, which indicates a gradual overall

intercycle softening of the material (Ng, McKinley, & Ewoldt, 2011) and reflects the overall behaviour

seen in fig 4a, top row. This behaviour is likely to be caused by alignment of network segments within

the flow field or loss of network junctions that – similar to the effects observed in thixotropy tests – may

lead to network segments which are unable to rejoin the network (Hyun et al., 2011). Moreover, in this

range of intercycle strain amplitude γ0 the shape of the Lissajous plots begins to get distorted from the

elliptical shape, and elastic stress τ’ starts to deviate from a straight line showing a clear impact of higher

harmonics. However, depending on the pH value during pre-treatment, the distortion to the Lissajous

plots is of different character and magnitude. If samples were pre-treated at pH 6.0 or pH 6.5, the

corresponding elastic plots (fig 4d, II) begin to rotate and distort from an elliptical shape sooner, but the

distortion appears to be gradual and fairly uniform. With increasing pH value during pre-treatment,

rotation starts at higher amplitudes, and – especially for samples pre-treated at pH 7.5 and pH 8.0 – the

distortion from elliptical shape starts more sudden (γ0 = 6.3%) and leads to more pronounced changes

in shape. Lissajous plots from samples pre-treated at pH 7.0 show intermediate rotation and deformation

behaviour. In addition, the shapes of the Lissajous plots are widening thus increasing the enclosed area

indicating an increasingly dissipative response (Ng et al., 2011). This effect is further quantified in the

dissipation ratio φ (fig 4a II second row) which starts to become dependent of the intercycle strain

amplitude γ0 and begins to increase, thus indicating a beginning shift from predominantly elastic

behaviour towards an increasing contribution of plastic properties (Ewoldt et al., 2010). The deviation

in the shape of Lissajous plots and the increase in dissipation ratio φ are accompanied by the transition

of elastic stress τ’ curves from a straight line towards an inversed sigmoidal shape. The decreased slope

at low intracycle strain γ indicates a beginning reduction of size and volume fraction of rigid clusters

(Park et al., 2015) and supports the beginning shift from elastic to plastic properties, while the increasing

slope at higher intracycle strain γ reflects the stretching of remaining clusters that cause intracycle

stiffening (Park et al., 2015). This is reflected in the increase of the S-factor (fig 4a II, third row) which

– in this range of intercycle stain amplitudes γ0 is likely to be related to intracycle strain stiffening since

the dissipation ratio φ is still well within the elastic range and the network structures only just began to

be disrupted.

At intercycle strain γ0 yet another order of magnitude higher (10% ≤ γ0 ≤63%, fig 4b and c III) all shapes

are distorted and show clear differences between samples. While Lissajous plots of samples pre-treated

at pH 6.0 or pH 6.5 assume slightly bone shaped profiles, pre-treatment at pH 8.0 leads to Lissajous

plots that are nearly rectangular shaped which implies an approximation towards perfect plastic

behaviour (fig 4b and c III) (Ewoldt et al., 2010). Samples pre-treated at pH-values in between, show

in-between distortions. In this range of intercycle strain amplitude γ0 the clockwise rotation also

proceeds, as visible in Lissajous plots, and – more distinctly – in the elastic stress (fig 4b III).

Consequently, the slopes at small intracycle strain γ decrease further, indicating a continued intercycle

strain softening behaviour owing to further reduction of rigid clusters. Nevertheless, at large intracycle

strain the slope of the elastic stress still increases apparently indicating some remaining intracycle strain

stiffening properties and therefore some residual elastic properties resulting in continuously positive S-

factors in fig 4a III (third row). Despite overall similarities of the elastic stress curves, differences

between samples become apparent in the considered range of intercycle strain γ0. While samples pre-

treated at pH 6.0 and pH 6.5 showed more moderate rotation and a less pronounced inversed sigmoidal

shape, increased pH values during pre-treatment led to an almost horizontal slope at low intracycle

strain γ which indicates an almost complete loss of rigid structures. These observations are reflected in

the development of the dissipation ratio φ (fig 4a III, second row). While samples pre-treated at pH 6.0

and pH 6.5 only reach values in the range of φ ≈ 0.5 in the discussed range of intracycle strain

amplitudes γ0 (10% ≤ γ0 ≤63) pre-treatment at pH 7.5 or pH 8.0 leads to φ > π/4 and therefore – in

agreement with the vast loss of rigid structures deduced from elastic stress τ’ and the shape of the

Lissajous plots – the intracycle behaviour becomes predominantly plastic for these samples.

At the highest intercycle strain amplitudes γ0 distortion of the Lissajous plots proceeds even further

(fig 4b IV). In case of samples pre-treated at pH 6.0 and pH 6.5, the increase in dissipated energy as

indicated by the increase of the enclosed area in Lissajous plots (fig 4b IV) is accompanied by a

transition from intracycle strain stiffening to intracycle strain softening behaviour as derived from the

shift of elastic stress τ’ curves from inversed sigmoidal to sigmoidal shape at the highest intercycle

strains γ0. This shift is accompanied by a decrease of the S-factor below 0 and a second increase in the

curves of the dissipation ratio φ. Despite the increase in dissipation ratio φ to final values of 0.64 ± 0.01

(pre-treatment at pH 6.0) and 0.73 ± 0.07 (pre-treatment at pH 6.5), these samples maintain a relevant

proportion of elastic behaviour within the investigated deformation range, since final values of the

dissipation ratio φ remain below π/4 and the S-factors become negative. In contrast, for samples pre-

treated at pH 7.0 to pH 8.0 the shapes approach rectangles with slightly rounded tops and bottoms which

are similar to those described by (Ewoldt et al., 2010) for predominantly plastic systems. The shift to

predominantly plastic behaviour is further reflected in the dissipation ratio φ that reaches final values of

0.84 ± 0.01, 0.88 ± 0.00 and 0.89 ± 0.00 for samples pre-treated at pH 7.0, pH 7.5 and pH 8.0

respectively and S-factors that increase further and approach final values of 0.66 ± 0.06, 0.87 ± 0.09 and

0.96 ± 0.03 respectively. Since positive S-factors generally indicate strain stiffening behaviour but may

also indicate a transition from elastic to plastic behaviour as explained in the materials and methods

section above, a closer investigation of the obtained S-factors is necessary. The curves of the S-factor

(fig 4a, bottom row) show an indentation at the beginning of the considered amplitude range (100%

≤ γ0 ≤ 630%, section IV) followed by a second increase. This indentation is most pronounced upon pre-

treatment at pH 7.0 and decreases with increasing pH during pre-treatment. While it is reasonable to

assume that in analogy to the curves of samples pre-treated at pH 6.0 and pH 6.5 the initial increase in

these curves can be related to intracycle strain stiffening effects, this second increase must not

necessarily be an indicator of continuing intracycle strain stiffening behaviour at the highest intercycle

strain amplitudes γ0. In fact, this would be very unlikely based on the presented results for elastic stress

at the corresponding intercycle strain amplitudes γ0 but can rather be explained by the fact, that in this

case the tangent modulus at minimum intracycle strain γ (GM) becomes negligible and the S-factor

consequently diverges towards one as explained in the materials and methods section above for plastic

behaviour (Ewoldt et al., 2010). Overall, these samples lost the majority of elastic properties while

samples pre-treated at pH 6.0 and pH 6.5 retained a higher proportion of elastic properties. This leads

to the conclusion, that coarser network structures are less prone to a complete transition towards plastic

properties under the applied oscillatory strain conditions owing to an increased structural flexibility.

The underlying effects and differences between rheological behaviour of the samples outside the linear

viscoelastic regime can best be related to their differences in network densities. In summary, all samples

show intercycle strain softening, as derived from the clockwise rotation of the Lissajous plots (Ng et al.,

2011) and elastic stress curves. For samples pre-treated at pH 6.0 and pH 6.5, the rotation starts sooner,

indicating a higher flexibility of coarser networks to follow deformation, as a more flexible network is

more likely to orient in the flow field. Additionally, the higher flexibility leads to a more gradual

decrease in bond numbers and is reflected in transition from intracycle strain stiffening towards

intracycle strain softening at higher intercycle strain γ0. Denser gels (i.e. samples pre-treated at pH 7.0

to 8.0) are able to resist intercycle strain softening slightly longer, but – owing to their denser structures

– have a higher degree of crosslinking and are therefore more prone to small microcracks (Faber et al.,

2017) and decreasing bond numbers (Hsiao et al., 2012; Park & Ahn, 2013; Park et al., 2015) in the gel

network. This in turn leads to the gradual disintegration of the gel and the transition from predominantly

elastic properties to mainly plastic behaviour.

CONCLUSIONS

In summary our results show, that different pH values during pre-treatment of pea protein lead to

different ratios between soluble and insoluble protein aggregates in the protein slurry before

fermentation. These different ratios in turn are the determining factor, when it comes to the degree of

crosslinking during gelation and the content of inactive fillers, that may disrupt the gel network structure

and thus directly influence the rheological properties. It was found, that all gels showed frequency

dependencies similar to milk yoghurts. However, storage and loss moduli were at significantly different

magnitudes if pre-treatment was carried out between pH 6.5 and pH 8.0, confirming different degrees

of crosslinking. Furthermore, the ability to recover network structures after intense shearing decreased

with decreasing pH values during pre-treatment. Large amplitude oscillatory shear rheology indicated

an overall intercycle strain softening behaviour and further confirmed differences in network densities.

While denser networks started to decompose into smaller clusters sooner and eventually changed from

predominantly elastic to mainly plastic behaviour, coarser networks displayed a higher flexibility

towards deformation. Adjusting the pH value during pre-treatment of pea protein prior to fermentation

induced gelation is therefore a valid tool for the customisation of the texture properties of pea protein-

based yoghurt alternatives. With this knowledge in mind – besides catering to consumers preferences

concerning yoghurt alternatives – further opportunities for the development of a wide range of fermented

pea protein products with different texture requirements such as bread spreads, confectionary fillings

etc. open up.

Acknowledgements:

The authors gratefully acknowledge the assistance and expertise of L. Barthel for CLSM measurements

and the skilful lab-work of C. Härter as well as the proofreading by M. Brückner-Gühmann.

Funding:

This work was supported by NutriAct – Competence Cluster Nutrition Research Berlin-Potsdam

funded by the Federal Ministry of Education and Research (BMBF) (FKZ: 01EA1806C).

References

Banovic, M., Lähteenmäki, L., Arvola, A., Pennanen, K., Duta, D. E., Brückner-Gühmann, M., &

Grunert, K. G. (2018). Foods with increased protein content: A qualitative study on European

consumer preferences and perceptions. Appetite, 125, 233–243.

https://doi.org/10.1016/j.appet.2018.01.034

Barac, M., Cabrilo, S., Pesic, M., Stanojevic, S., Zilic, S., Macej, O., & Ristic, N. (2010). Profile and

functional properties of seed proteins from six pea (Pisum sativum) genotypes. International

Journal of Molecular Sciences, 11(12), 4973–4990. https://doi.org/10.3390/ijms11124973

Britten, M., & Giroux, H. J. (2001). Acid-induced gelation of whey protein polymers: Effects of pH

and calcium concentration during polymerization. Food Hydrocolloids, 15(4–6), 609–617.

https://doi.org/10.1016/S0268-005X(01)00049-2

Cairoli, S., Iametti, S., & Bonomi, F. (1994). Reversible and irreversible modifications of β-

lactoglobulin upon exposure to heat. Journal of Protein Chemistry, 13(3), 347–354.

https://doi.org/10.1007/BF01901568

Cano-Sarmiento, C., Téllez-Medina, D. I., Viveros-Contreras, R., Cornejo-Mazón, M., Figueroa-

Hernández, C. Y., García-Armenta, E., … Gutiérrez-López, G. F. (2018). Zeta Potential of Food

Matrices. Food Engineering Reviews, 10(3), 113–138. https://doi.org/10.1007/s12393-018-9176-

Crévieu, I., Carré, B., Chagneau, A.-M., Quillien, L., Guéguen, J., & Bérot, S. (1997). Identification of

Resistant Pea ( Pisum sativum L.) Proteins in the Digestive Tract of Chickens. Journal of

Agricultural and Food Chemistry, 45(4), 1295–1300. https://doi.org/10.1021/jf960806b

Croy, R., Derbyshire, E., Krishna, T., & Boulter, D. (1979). Legumin of pisum sativum and vicia faba.

The New Phytologist, 29–35.

Croy, R., Gatehouse, J., Evans, I., & Boulter, D. (1980). Characterisation of the storage protein

subunits synthesised in vitro by polyribosomes and RNA from developing pea (Pisum sativum

L.) - I. Legumin. Planta, 148(1), 49–56. https://doi.org/10.1007/BF00385441

Croy, R., Gatehouse, J., Tyler, M., & Boulter, D. (1980). The purification and characterization of a

third storage protein (convicilin) from the seeds of pea ( Pisum sativum L.). Biochemical

Journal, 191(2), 509–516. https://doi.org/10.1042/bj1910509

Damodaran, S., & Kinsella, J. E. (1982). Effect of Conglycinin on the Thermal Aggregation of

Glycinin. Journal of Agricultural and Food Chemistry, 30(5), 812–817.

https://doi.org/10.1021/jf00113a003

Dickinson, E. (2009). Hydrocolloids as emulsifiers and emulsion stabilizers. Food Hydrocolloids,

23(6), 1473–1482. https://doi.org/10.1016/j.foodhyd.2008.08.005

Dziuba, J., Szerszunowicz, I., Nalecz, D., & Dziuba, M. (2014). Proteomic Analysis of Albumin and

Globulin Fractions of Pea ( Pisum Sativum L .) Seeds *. Acta Scientarium Polonorum, 13, 181–

190.

Ewoldt, R. H., Hosoi, A. E., & McKinley, G. H. (2008). New measures for characterizing nonlinear

viscoelasticity in large amplitude oscillatory shear. Journal of Rheology, 52(6), 1427–1458.

https://doi.org/10.1122/1.2970095

Ewoldt, R. H., Winter, P., Maxey, J., & McKinley, G. H. (2010). Large amplitude oscillatory shear of

pseudoplastic and elastoviscoplastic materials. Rheologica Acta, 49(2), 191–212.

https://doi.org/10.1007/s00397-009-0403-7

Faber, T. J., Van Breemen, L. C. A., & McKinley, G. H. (2017). From firm to fluid – Structure-texture

relations of filled gels probed under Large Amplitude Oscillatory Shear. Journal of Food

Engineering, 210, 1–18. https://doi.org/10.1016/j.jfoodeng.2017.03.028

Gatehouse, J., Croy, R., Morton, H., Tyler, M., & Boulter, D. (1981). Characterisation and Subunit

Structures of the Vicilin Storage Proteins of Pea (Pisum sativum L.). European Journal of

Biochemistry, 118(3), 627–633. https://doi.org/10.1111/j.1432-1033.1981.tb05565.x

Gatehouse, J., Lycett, G., Croy, R., & Boulter, D. (1982). The post-translational proteolysis of the

subunits of vicilin from pea ( Pisum sativum L.). Biochemical Journal, 207(3), 629–632.

https://doi.org/10.1042/bj2070629

Gatehouse, J., Lycett, G., Delauney, A., Croy, R., & Boulter, D. (1983). Sequence specificity of the

post- translational proteolytic cleavage of vicilin, a seed storage protein of pea (Pisum sativum

L.). Biochemical Journal, 212, 427–432.

German, B., Damodaran, S., & Kinsella, J. E. (1982). Thermal Dissociation and Association Behavior

of Soy Proteins. Journal of Agricultural and Food Chemistry, 30(5), 807–811.

https://doi.org/10.1021/jf00113a002

Guzmán-González, M., Morais, F., Ramos, M., & Amigo, L. (1999). Influence of skimmed milk

concentrate replacement by dry dairy products in a low fat set-type yoghurt model system. I: Use

of whey protein concentrates, milk protein concentrates and skimmed milk powder. Journal of

the Science of Food and Agriculture, 79(8), 1117–1122. https://doi.org/10.1002/(SICI)1097-

0010(199906)79:8<1117::AID-JSFA335>3.3.CO;2-6

Hassan, A. N., Ipsen, R., Janzen, T., & Qvist, K. B. (2003). Microstructure and rheology of yogurt

made with cultures differing only in their ability to produce exopolysaccharides. Journal of

Dairy Science, 86(5), 1632–1638. https://doi.org/10.3168/jds.S0022-0302(03)73748-5

Hsiao, L. C., Newman, R. S., Glotzer, S. C., & Solomon, M. J. (2012). Role of isostaticity and load-

bearing microstructure in the elasticity of yielded colloidal gels. Proceedings of the National

Academy of Sciences of the United States of America, 109(40), 16029–16034.

https://doi.org/10.1073/pnas.1206742109

Hyun, K., Wilhelm, M., Klein, C., Cho, K., Nam, J., Ahn, K., … McKinley, G. (2011). A review of

nonlinear oscillatory shear tests: Analysis and application of large amplitude oscillatory shear

(LAOS). Progress in Polymer Science (Oxford), 36(12), 1697–1753.

https://doi.org/10.1016/j.progpolymsci.2011.02.002

Klemmer, K. J., Waldner, L., Stone, A., Low, N. H., & Nickerson, M. T. (2012). Complex

coacervation of pea protein isolate and alginate polysaccharides. Food Chemistry, 130(3), 710–

715. https://doi.org/10.1016/j.foodchem.2011.07.114

Klost, M., & Drusch, S. (2019). Structure formation and rheological properties of pea protein-based

gels. Food Hydrocolloids, 94(September), 622–630.

https://doi.org/10.1016/j.foodhyd.2019.03.030

Klost, M., Giménez-Ribes, G., & Drusch, S. (2020). Enzymatic hydrolysis of pea protein: Interactions

and protein fractions involved in fermentation induced gels and their influence on rheological

properties. Food Hydrocolloids, 105, 105793. https://doi.org/10.1016/j.foodhyd.2020.105793

Krajcovicova-Kudlackova, M., Babinska, K., & Valachovicova, M. (2005). Health benefits and risks

of plant proteins. Bratislavské Lekárske Listy, 106(6–7), 231–234.

Mermet-Guyennet, M. R. B., Gianfelice de Castro, J., Habibi, M., Martzel, N., Denn, M. M., & Bonn,

D. (2015). LAOS: The strain softening/strain hardening paradox. Journal of Rheology, 59(1),

21–32. https://doi.org/10.1122/1.4902000

Mezger, T. (2006). Das Rheologie Handbuch (2nd ed.). Hannover: Vincentz Network GmbH & Co.

KG.

Ng, T. S. K., McKinley, G. H., & Ewoldt, R. H. (2011). Large amplitude oscillatory shear flow of

gluten dough: A model power-law gel. Journal of Rheology, 55(3), 627–654.

https://doi.org/10.1122/1.3570340

Nishinari, K., Fang, Y., Guo, S., & Phillips, G. O. (2014). Soy proteins: A review on composition,

aggregation and emulsification. Food Hydrocolloids, 39, 301–318.

https://doi.org/10.1016/j.foodhyd.2014.01.013

Park, J. D., & Ahn, K. H. (2013). Structural evolution of colloidal gels at intermediate volume fraction

under start-up of shear flow. Soft Matter, 9(48), 11650–11662.

https://doi.org/10.1039/c3sm52090k

Park, J. D., Ahn, K. H., & Lee, S. J. (2015). Structural change and dynamics of colloidal gels under

oscillatory shear flow. Soft Matter, 11(48), 9262–9272. https://doi.org/10.1039/c5sm01651g

Petruccelli, S., & Añón, M. C. (1995). Thermal Aggregation of Soy Protein Isolates. Journal of

Agricultural and Food Chemistry, 43(12), 3035–3041. https://doi.org/10.1021/jf00060a009

Piorkowski, D. T., & McClements, D. J. (2014). Beverage emulsions: Recent developments in

formulation, production, and applications. Food Hydrocolloids, 42, 5–41.

https://doi.org/10.1016/j.foodhyd.2013.07.009

Ringgenberg, E., Alexander, M., & Corredig, M. (2013). Effect of concentration and incubation

temperature on the acid induced aggregation of soymilk. Food Hydrocolloids, 30(1), 463–469.

https://doi.org/10.1016/j.foodhyd.2012.05.011

Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image

analysis. Nature America, 9, 671–675.

Swanson, B. G. (1990). Pea and Lentil Protein Extraction and Functionality. Journal of the American

Oil Chemists’ Society, 67(5), 276–280.

WHO. (2013). Draft action plan for the prevention and control of noncommunicable diseases 2013 –

2020. Report by the Secretariat, (April), 1–50.

Yamagishi, T., Miyakawa, A., Noda, N., & Yamauchi, F. (1983). Isolation and Electrophoretic

Analysis of Heat-induced Products of Mixed Soybean 7S and 11S Globulins. Agricultural and

Biological Chemistry, 47(6), 1229–1237. https://doi.org/10.1271/bbb1961.47.1229

SUPPLEMENTARY

Fig S 1 Molecular weight profile of pea protein (black) and pea protein hydrolysed with Protamex® under different pH

conditions (blue: pH 8.0, grey: pH 7.5, red: pH 7.0) and storage moduli G’ of the corresponding fermentation induced gels.

Fig S2 pH-drop (a) and increase of G’ (b) during fermentation of pea protein preheated at different pH-values.

Fig S3 G’ and G’’ in frequency sweeps pH 6.0 (a), pH 6.5 (b), pH 7.0 (c), pH 7.5 (d), pH 8.0 (e)