Investigating the origin of bitter peptides in soy protein extracts using LC-MS based proteomics

(1)

MSc Chemistry

Analytical Chemistry

Master Thesis

Investigating the origin of bitter peptides in soy protein

extracts using LC-MS based proteomics

by

Lars Bouhuijs

12435589

May 2020

48EC

May 2020 - April 2021

Supervisor/Examiner:

Examiner:

dr. A. Gargano

prof. dr. G.L. Corthals

Department: HIMS in cooperation

with Unilever

(2)

2

Abstract

The production of meat leads to the emission of nitrogen and greenhouse gasses, which affect the ecosystem negatively. Also, the consumption of saturated fat is linked to an increased risk of cardiovascular diseases and strokes, while the intake of red meat is associated with colorectal cancer. The production and consumption of plant-based meat alternatives could solve these problems. However, during the production of plant-based meat alternatives, bitter peptides are formed and these give an unpleasant taste to the meat alternative. Soy is one of the plant-based products used in meat alternatives. Therefore, the origin of the bitter peptides in soy protein powder has been investigated.

First, an extraction method with acetone precipitation has been developed to extract proteins. This has been assessed by SDS-PAGE and works for extracting proteins. Next, a denaturing size-exclusion chromatography method has been developed to separate the proteins from the peptides by fractionating the samples. The peptide fractions have been subjected to peptide analysis. The bitterness of the identified peptides has been calculated by iBitter-SCM [1]. Soy 6, a hydrolysed soy protein, was found to contain more bitter peptides and the bitter peptides also contained higher intensities.

The protein fractions are split into two. This way different methods can be applied to proteins and peptides. One part is used for a bottom-up approach to identify the proteins present. All of the high abundant proteins such as glycinin, β-conglycinin, 2S albumin and 7S globulin have been identified in the bottom-up approach. Of these proteins, glycinin and 2S albumin fragments have been identified during the top-down analysis. No proteins larger than 35 kDa have been identified. Furthermore, a fragment of Bowman-Birk type proteinase inhibitor has been identified during the top-down analysis. Because the protein is no longer intact, it is expected that it is no longer enzymatically active. However, the top-down method is not optimal for proteins above 35 kDa and thus needs further research. If this method is optimised it can be used to analyse larger proteins, such as lipoxygenases on their intactness. Furthermore, the method can be used to predict the origin is of bitter peptides in meat alternatives by comparing the raw materials with the meat alternative.

Thousands of potential bitter peptides have been identified in two soy samples. For several of them, glycinin and β-conglycinin were identified as potential origins of the bitter peptides. Furthermore, uncharacterised proteins were also identified as potential origins.

(3)

3

List of abbreviations

AA Amino acid

CID Collision-induced dissociation Da Dalton

DDA Data-dependent acquisition DH Degree of hydrolysis

DIA Data-independent acquisition ESI Electrospray ionization

F1, F2, F3 etc. Fraction 1, Fraction 2, Fraction 3, etc HCD Higher-energy collisional dissociation HPLC High-performance liquid chromatography

LC-MS/MS Liquid chromatography-tandem mass spectrometry m/z Mass-to-charge ratio

MALDI Matrix-assisted laser desorption-ionization NaCaHs Sodium caseinate protein hydrolysates

ppm Parts per million

PTMs Post-translational modifications

QSAR Quantitative structure-activity relationship RP Reverse-phase

rpm Revolutions per minute S3 Soy 3

S6 Soy 6

SCM Scorecard method

SDS-PAGE Sodium dodecyl sulphate-polyacrylamide gel electrophoresis SEC Size exclusion chromatography

SIM Single ion monitoring SPE Solid-phase extraction TCA Trichloroacetic acid TUV Tuneable ultra-violet

(4)

4

1. Introduction

Meat has been regarded as an essential part of a healthy diet, mainly because it is one of the main sources of proteins besides dairy, nuts, legumes, and grains [2]. In 2011, about half of the protein intake in the Netherlands originated from animal sources [3]. However, the production of meat leads to the emission of nitrogen and greenhouse gasses, which affect the ecosystem negatively [2,4]. With the development of meat alternatives, the consumption of meat can be decreased, influencing the ecosystem positively. Meat alternatives are foods that do not contain meat, poultry, fish or shellfish and that have similar taste, appearance, nutritional value, and texture to food made from meat, poultry, fish or shellfish [5]. Westhoek et al. (2014) [4] expects that when meat, dairy, and egg consumption is halved, the emission of nitrogen and greenhouse gas is reduced by 40% and 25-40%, respectively. Furthermore, the changes in diet would also have health benefits. There would be a reduction in the intake of saturated fat and red meat. The intake of saturated fat is linked to an increased risk of cardiovascular diseases and strokes, while the intake of red meat is associated with colorectal cancer [4,6]. Because of all these benefits, the popularity of meat alternatives is increasing. Currently, there are two kinds of meat alternatives developed: cultured meat which is produced from cells, and plant-based meat alternatives [2]. Nowadays, plant-based meat alternatives are often produced from protein isolate and concentrate from soy, pulse seeds, and potato. During the production of plant-based meat alternatives, the proteins present in the plants are hydrolysed to peptides [7,8]. A big downside of this process is that often bitter peptides are formed and these bitter peptides are unwanted. Bitter taste aversion is innate because the taste is linked to bitter-tasting toxic substances [9]. Hydrolysis can occur under different circumstances. For example, the addition of an enzyme can cleave the proteins into peptides [8,10]. Different enzymes cleave at different amino acids (AAs) causing different peptide lengths and different end groups. The peptide length and end groups influence the bitterness but also the hydrophobicity determine how bitter a peptide is. Bitter peptides are often small peptides. Therefore, the degree of hydrolysis (DH) is related to bitterness. As a result, the bitterness of the peptides formed is different for each enzyme used.

Several studies have been done on the identification of bitter peptides [7,11–13]. Most of the studies combine identification with sensory analysis to identify the bitter peptides [12–14]. However, the bitterness can also be determined by calculation after identification with the use of quantitative structure-activity relationship (QSAR) studies, or scoring card methods with propensity scores of dipeptides for example [1,15,16]. The last way to determine the bitterness is with the use of an electronic tongue [14]. However, this method is not as precise as a sensory panel. As discussed, a lot of research has been done on the identification of bitter peptides but very little research has been done on the intact protein that forms the bitter peptides. Another group of proteins that are of interest to the food industry are Bowman-Birk proteinase inhibitors and lipoxygenases. Lipoxygenases can form oxidation products of polyunsaturated fatty acids and those products have a beany and grassy flavour and aroma and this is unwanted in meat alternatives [17,18].

In this present work, the origin of bitter peptides in soy protein extracts has been elucidated by an intact protein and bottom-up liquid chromatography-orbitrap tandem mass spectrometry (LC-MS/MS) approach. Furthermore, this approach can be used to distinguish differences between the raw

(7)

7

materials and the processed meat alternative at a protein and peptide level. First, proteins are extracted from soy protein powder. The applied extraction method has been assessed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE). Second, denaturing size exclusion chromatography (SEC) has been applied to separate the naturally occurring peptides from the proteins in soy powders by fractionation. The working of the SEC fractionation has been verified by SDS-PAGE. After fractionation, the intact method has been applied to the collected fractions to determine if the proteins are still intact and if there are any post-translational modifications (PTMs) present. Furthermore, a bottom-up approach has been done to identify the proteins present in different fractions. This information has been coupled to the intact analysis of the fractions to know what proteins should be present and if these proteins are still intact. Last, the bitter peptides have been identified in different soy protein powders by LC-MS/MS.

(8)

8

2. Background

Several interesting points such as what causes bitterness, the way of determining the bitterness and how bitter peptides are formed have been shortly mentioned in the introduction. In this chapter, more detailed information is given on what characteristics cause bitterness, how the bitter peptides are formed, how the bitterness can be determined, and which proteins are present in soy. During this study, the main focus will be on soy, even though meat alternatives often contain more and other plant-based protein sources.

2.1. What causes bitterness?

There are several peptide characteristics responsible for bitterness. The first characteristic that is responsible for bitterness is hydrophobicity [9,19,20]. This has been proven in the 1970s by Ney (1979) [19]. The bitter peptides need to be hydrophobic to be able to have interaction with the bitter receptors (T2Rs) on the tongue to taste bitterness [9]. Second, the end groups influence the bitterness. For example, a hydrophobic C-terminal amino acid (bulky for peptides ≥ tetrapeptides) and a bulky positively charged hydrophobic N-terminal amino acid can greatly increase the bitterness of a peptide. A bulky amino acid that is often present at the terminals is leucine. Another amino acid that can increase the bitterness is arginine. At the N-terminal, arginine greatly increases the bitterness, while it does not increase the bitterness as much at the C-terminal. However, when the end groups stay the same, the order of the amino acids in the amino acid sequence does not greatly influence the bitterness [16]. Furthermore, bitter peptides are often smaller peptides, from two to ten amino acids [9,15,20]. Additionally, post-translational modifications (PTMs) can influence bitterness. For example, when peptides have acetylation’s and/or esterification’s at the amino and/or carboxyl group the bitterness increased ten-fold [9]. When both the end groups are blocked by the acetylation and/or esterification the bitterness is greater than when only 1 group is blocked. On the other side, glycation can decrease bitterness because the attachment of sugars decreases the hydrophobicity of the peptides [21]. Last, the concentration influences bitterness. When the quantity of a bitter peptide increases, the bitterness become more intense.

2.2. How are bitter peptides formed?

Bitterness is related to the degree of hydrolysis (DH) [10]. The DH is defined as the fraction of cleaved peptide bonds in a protein hydrolysate [22]. When the DH increased, the bitterness increased when several different proteases were used. The increase in bitterness can be linked to the peptide chain length. The more peptides are formed from a protein, the smaller the peptides are and as mentioned the chain length influences the bitterness. Furthermore, the kind of protease used affects the bitterness. For example, when alcalase was used during hydrolysis of soy proteins, the highest bitterness was obtained, whereas neutrase and trypsin resulted in a lower bitterness [23]. This can be explained by the different end groups present in the cleaved peptides. Every enzyme cleaves at certain amino acids, resulting in different end groups but also different chain lengths because certain amino acids are more common than others.

(9)

9

Seo et al. (2008) [10] studied several proteases and the corresponding bitterness of the hydrolysis. Alcalase showed again the highest bitterness and flavourzyme showed the lowest bitterness. Flavourzyme is a mixture of endoprotease and exopeptidase, while the other proteases were endoprotease. Endoproteases tend to hydrolyse at hydrophobic amino acid residues. As a result, the peptide is left with a hydrophobic C- or N-terminal amino acid. Exopeptidases often hydrolyse a single amino acid from the end of the peptide, thus removing the hydrophobic amino acid and reducing the bitterness [24]. Meinlschmidt et al. (2016) [8] also studied the influence of several proteases, such as alcalase, flavourzyme, and papain (cysteine-protease) on the bitterness of SPI. Alcalase gave the worst results but flavourzyme and especially papain gave the best results. Another study showed contradictory results [25]. Alcalase did show the highest bitterness again but flavourzyme also showed a high bitterness. In this study, the lowest bitterness was achieved when papain and α-chymotrypsin were used during hydrolysis. However, this was for pea protein hydrolysates and not for soy as the two studies mentioned above.

Looking at the literature discussed above. It seems likely that the bitter peptides are mainly formed due to the industrial process of adding a proteasome and hydrolysing the protein. This can be concluded from the hydrolysis with flavourzyme where little bitter peptides were formed. However, bitter peptides are also detected in fermented foods, where no enzymes are added but the process produces bitter peptides [9].

Hydrolysis also occurs at a rapid rate when the temperature is increased [26]. This may be caused by the thermal denaturation of the protein. As a result, the cleavage sites are more accessible. This is especially of interest for the preparation of meat alternatives because thermal extrusion is the predominant technique used during the process of making meat alternatives [27]. During extrusion, the proteins are heated and extruded. When looking at different hydrolysis conditions and different proteasomes to reduce the bitterness, a proteasome with an exoprotease and an endoprotease would be optimal to reduce the bitterness of the formed peptides. However, the most optimal would be to minimise the hydrolysis of the proteins. As a result, fewer bitter peptides would be formed.

2.3. How can bitterness be determined?

The bitterness can be determined in multiple ways. An overview of the possibilities is given in Table 1. The first and most obvious way to determine the bitterness is by tasting, also called a sensory test [10,13,20,28,29]. During these studies, a panel of trained assessors is presented with a triangle test of 1 bitter sample and 2 blanks. Caffeine standards are often used as reference standards to train the assessors. The assessors will then rate the bitterness of the sample on a scale of 1 to 10, corresponding to the bitterness of the caffeine standards. This is the most trustworthy way of determining bitterness, even though not every human has the same taste and human error is possible.

Another way to determine the bitterness is with the use of an electronic tongue [14]. This is a taste sensing system and it is equipped with a lipid membrane sensor. The sensor represents the gustatory stimuli for bitterness. It showed good results in measuring the bitterness of sodium caseinate protein hydrolysates (NaCaHs) and the reduction of bitterness by sweeteners [30]. However, when the

(10)

10

bitterness of NaCaHs was measured in beverages, the results did not correspond to the sensory panel, showing that the electronic tongue is not optimal yet.

Another obvious way is to identify the peptide and search in a database with known bitter peptides. One of these databases is the BIOPEP [31] sensory database. In this database, known sensory peptides from previous studies are displayed. However, the downside of a database is that only known bitter peptides are presented. When an unknown bitter peptide has been found, it can be classified as a non-bitter because there is no match.

The last way to determine the bitterness is by calculating the bitterness according to the structure of the peptide [1,15,16,32]. These methods include QSAR studies and score card methods (SCM). One of the first ways to calculate bitterness is described by Ney (1979) [19]. According to Ney (1979) [19], bitterness is related to the average hydrophobicity of the peptide [24]. The hydrophobicity of a peptide (Q) is defined as the sum of free energies of transfer of the amino acid side chains from ethanol to water. The bitterness can be calculated with the following formula: 𝑄 = ∑ 𝛥𝑔/𝑛, where 𝛥𝑔 is the free energy transfer and n the number of amino acids. However, in several studies, when results of sensory analysis where compared with Q-values, they were contradictory [9,20,33]. The Q-value is thus not fully reliable to use for bitterness assessment. This can also be linked to what causes bitterness because also the end group has influence and this has not been taken into account in the Q-value.

When only the average hydrophobicity is taken into account, it can be unprecise [20]. As a result, many studies have been done and models have been developed to calculate the bitterness. One of the studies used a three z-score approach (log mass values, total hydrophobicity, and residue number) to determine the bitterness [15,34]. The z-scores have been determined in the 1980s by Hellberg et al. (1987) [34].

Multiple QSAR models have been developed and during these QSAR’s often multiple peptide characteristics are taken into account to gain a precise model. For example, a QSAR model with fourteen descriptor sets has been developed by Xu and Chung (2019) [32], while Soltani et al. (2013) [16] developed a model with six descriptors. These descriptors describe functions such as size, dimension, amount of atoms, polarizability, conformational properties, and electrical properties. The last way to calculate the bitterness is based on an SCM, called iBitter-SCM [1]. For the development of this method, known bitter peptides have been obtained from literature together with non-bitter peptides generated from the before mentioned BIOPEP database. From the collected peptides a training set and an independent set were made. The model was developed from the training set using a statistical approach and a genetic algorithm to optimize the initial score card and determine the optimal threshold value. A downside from such methods is that you gain a value and this value is above or below a cut-off value but it does not take concentration or amount into account. This can give problems for determining if a peptide is bitter. Nevertheless, it is a good indication of how bitter a peptide can be.

(11)

11 Table 1 An overview of the different ways of determining the bitterness

Method How is the bitterness determined? Source

Sensory test

Tasting the peptides The peptides are tasted by a trained panel and compared to caffeine to determine the bitterness.

[10,13,20,28,29]

Determining it with a tasting device

Electronic tongue A taste sensing system equipped with a lipid membrane sensor representing the gustatory stimuli for bitterness.

[14,30]

Database search Search for known bitter

peptides

The known bitter peptides are placed into a database that can be searched to identify the bitter peptides in the samples.

[31]

Calculation Calculating the

hydrophobicity

The first way to calculate the bitterness was by calculating the hydrophobicity because it is linked to bitterness.

[19]

Z-Score The bitterness can also be determined on more parameters than just hydrophobicity. With the z-score, bitterness is determined by the log mass values, total hydrophobicity, and residue number.

[15,34]

QSAR In a QSAR, the bitterness is calculated based on even more descriptors than with the z-score.

[16,32]

SCM This method is developed by using known bitter peptides and non-bitter peptides and this way creating a training set for an algorithm.

[1]

2.4. What proteins are present in meat alternatives?

As mentioned in the introduction, there are currently two kinds of meat alternatives: cultured meat and based meat [2,35]. During this study, only based meats are of interest. These plant-based meats consist of proteins and peptides of several plant origins such as soy, wheat, pea, and pulses. During this project, the focus is on proteins from soy.

Several studies have been done on identifying the proteins present in soybeans [7,36–38]. However, when the bitter peptides were determined, most of the time they were not linked to the corresponding protein or how they were formed. Therefore, gaining knowledge about the intact proteins during this study can be useful because it can be estimated which bitter peptides are formed and proteins can be compared before and after the preparation of the meat alternative. As a result, it can be determined

(12)

12

if bitter peptides are formed during the preparation of meat alternatives or if they are present in the raw materials.

For soybeans, 70-80% of the proteins are the storage proteins 7S (conglycinin) and 11S (glycinin). Additionally, it consists of 10% albumins and 20% oleosins. Because these proteins are the most abundant ones, it is expected that when bitterness is tasted, it originates from more abundant proteins since bitterness is also related to concentration. However, it does not mean that bitter peptides only originate from highly abundant proteins. During the protein identification of soy proteins, the lower abundant proteins are often not mentioned or found. Besides the storage proteins, lipoxygenases are also present in soy and lipoxygenases cause oxidation of polyunsaturated fatty acids. The products formed by lipoxygenases activity may yield products that have a beany and grassy flavour and aroma, which is unwanted in meat alternatives.

Besides soy, pulse seeds are also a source of high protein content and often used in plant-based meat alternatives [2]. The main proteins present in pulse seeds are storage proteins such as vicilin and

convicilin [38–40]. Additionally, there are also less abundant proteins such as prolamin, glutelin, trypsin-inhibitors, α legumin, and β legumin present. Storage proteins are often low in

sulphur-containing amino acids. α legumin and β legumin can make up for this since these are a source of sulphur-containing amino acids.

Something worth noting is that many of these plant-based products contain protease-inhibitors such as Kunitz trypsin-inhibitors and Bowman-Birk inhibitors. These inhibitors can decrease the efficiency of enzymes such as trypsin when used for hydrolysis. However, the inhibitors could be deactivated by thermal treatment. During this study, more proteins will be identified, in particular when they have PTMs or when they are truncated.

2.5. How are soy proteins extracted and processed?

There are different protein products made from soy [41]. The highest concentration proteins products are defatted soybean flakes/flour, soy protein concentrate (SPC), and soy protein isolate (SPI). Defatted soybean flour is the least processed soybean products, while soy protein concentrate and isolate are processed. Defatted soy flour has a protein content of about 50% and is produced by grinding defatted soy flakes. Soybean concentrate and isolate are higher in protein concentration and are produced by fractionating the defatted soy flakes. The concentrate has a protein content of about 70% and is produced by aqueous alcohol extraction. The isolate is produced by alkaline extraction followed by precipitation in acidic pH resulting in a protein content of 90%. This process is shown in Figure 1.

(13)

13

Figure 1 Overview of the different processes that are applied to soybeans to produce high protein content products. One of the ways of producing fibrous and meat-like textures is the use of shear structuring [42]. During this process, SPI is combined with wheat gluten (WG) and sodium chloride and subjected to shear at a temperature of 95 °C. The combination of heating and shearing the protein mix forms an anisotropic structure, which is preferred for meat alternatives. Thus, from soybean to potential meat alternative, the proteins go through different processing steps which could influence the structure of the proteins.

2.6. Analytical techniques

During this project, several analytical techniques will be used to analyse peptides and intact proteins. Size exclusion chromatography (SEC) will be used to separate the intact proteins from the peptides and thereby reducing sample complexity. Additionally, different separation methods can be applied for peptides and proteins. For example, for peptides a C18 column and for proteins a C4 column is used to separate the analytes before MS analysis. Therefore, it is beneficial to separate the proteins from the peptides. Furthermore, by applying SEC, the food stuff can be largely removed which also reduces sample complexity. After separating the proteins and peptides, a suitable LC-MS method can be applied.

2.6.1. Size exclusion chromatography

SEC is the technique of separating based on size, which is often linked to mass [43]. It is also called gel filtration, gel permeation or molecular sieve chromatography and can be used to fractionate and characterize proteins. The SEC column contains porous particles and molecules with a smaller diameter can enter the pores and move through the column more slowly. If the molecules are small enough to permeate through all the pores, they elute at the last point possible, also called the total permeation limit. Molecules with a larger diameter than the pores are excluded from the pores and are thus not retained in the column. If molecules are completely excluded, they elute at the so-called exclusion limit [44]. The speed of movement of the molecules and thus the separation is based on how easily the molecules can enter the pores. Therefore, the correct pore size must be chosen. However, the effective pore size can also be influenced by the solvents that are used [45]. For example, the addition of organic modifiers can decrease the effective pore size because of preferential solvation of the stationary phase and swelling of the stationary phase [46]. Additionally, the addition of organic modifiers to the eluent denatures the proteins. When proteins are denatured, they are fully or partially unfolded and thus the shape changes, which influences the retention in SEC.

(14)

14

SEC is often used for the fractionation of complex protein samples, as is the case for food samples [47,48]. It can be used to separate proteins from peptides and possibly remove the matrix. By fractionating the protein samples before LC-MS analysis, the sample complexity is reduced. However, the downside of conventional SEC is the use of salts. Salts are used to reduce interactions with the column, which are unwanted in SEC but salts are also unwanted in mass spectrometry [49]. The use of salts in SEC would require an extra desalting step. A way to overcome the problem of salts is to do denaturing SEC [50,51]. During denaturing SEC, acetonitrile and TFA are used to reduce the interaction with the column. This would overcome the problem of desalting, and thus the samples could be analysed straight away by LC-MS after concentrating. However, as mentioned above, the addition of an organic modifier causes the pore size to decrease and the proteins to denature.

2.6.2. Mass spectrometry

During this project, a Q Exactive Plus Orbitrap mass spectrometer (Figure 2) will be coupled to a nano-LC. The Orbitrap is one of the most high-end mass spectrometers, has high resolution and accurate mass detection [52,53]. Therefore, the use in proteomics has rapidly grown over the past years. The Q Exactive Plus Orbitrap is an orbitrap combined with a quadrupole. The advantage of the combination of a quadrupole with an Orbitrap is that the quadrupole can select ions almost instantaneously and in the higher-energy collisional dissociation (HCD) the ions are fragmented in a similar time scale. HCD operates at higher collision energies than collision-induced dissociation (CID) [54]. For peptide analysis, HCD mainly produces y-ions and to a lesser extent shorter b-ions. The HCD can be operated at different energies [54,55]. As a result, different fragmentations occur and this can help in identifying peptides, possibly also short peptides (<5 AAs). Short peptides are often hard to identify because they do not have favourable charges, they are suppressed by more abundant and larger peptides, and the fragmentation spectra cannot give sufficient information to identify the peptide [56,57]. The use of different collision energies or stepped collision energies can help improve the identification of short peptides [54].

The quadrupole mass filter also enables multiplexed scan modes such as selected ion monitoring (SIM) [52,53]. During SIM the quadrupole rapidly switches between selected m/z to pass into the C-trap where they are accumulated. Once the desired amount of ions are accumulated they are sent into the Orbitrap for a single detection event. The desired amount of ions is also called the AGC target. A benefit of applying SIM is that it is more sensitive because fewer m/z have to be scanned. However, to apply SIM, the composition of the sample needs to be known, which does not apply to the food samples measured during this project. However, it could be used to look more into detail at specific proteins or peptides of interest.

Because SIM is not a possibility, data-dependent acquisition (DDA) and data-independent acquisition (DIA) is widely used in bottom-up applications [58]. DIA analyses everything in a certain mass range, whereas DDA selects a precursor during MS1 based on the intensity or sets parameters and analyses it further in MS2. The benefits of DDA is that the analysis time is used effectively and the collection of uninformative MS2 spectra is minimized. However, a downside is that it does only collect the most abundant ions and could thus miss less abundant ions. This problem can be solved by using an

(15)

15

exclusion list and thus excluding the most abundant ions. Additionally, an inclusion list can be used to target specific proteins and peptides. However, this does make it a targeted approach. More targeted methods are selected reaction monitoring (SRM) and parallel reaction monitoring (PRM). In SRM, a precursor ion is selected and fragmented but only one or a few product ions are analysed. This is a very reproducible way of detecting and quantifying known peptides. PRM works in a very similar way, only every product ion is analysed. As a result, a greater number of transitions are shown to confirm the identity of the peptides. More methods for specifically analysing intact proteins are discussed in the next chapter.

Electrospray ionization (ESI) is the most applied ionization technique during top-down analysis [48,59– 62]. It is often applied over the soft ionization technique matrix-assisted laser desorption-ionization (MALDI) because the use of ESI results in more charges per protein. As a result, the mass-to-charge (m/z) ratio becomes lower and this facilitates the gas-phase fragmentation and therefore, the characterisation of the primary sequence and possible PTMs. Furthermore, ESI can be coupled to the LC. Several points have to be taken into account to improve ionization and decrease signal suppression. The first point is the formation of desolvated protein ions. Substances such as salts, detergents, chaotropes, and buffers can interfere in this process and lead to signal suppression. They often interfere after the formation of nanodroplets at the Rayleigh charge limit. Two processes occur in these droplets, partitioning of the net charge towards the surface of the droplet and the minimization of solvation energy [59]. Polar components such as salts and native proteins partition towards the centre of the droplet to optimize the solvation energy, resulting in a need to evaporate the solvent molecules. Hydrophobic components such as denatured proteins and detergents migrate to the surface of the droplet to optimize the solvation energy, resulting in a process that requires less energy, evaporation or ejection. The second point is the distribution of the MS signal across multiple channels, which increases with protein size. Each channel has its noise and when the signal is spread over multiple channels the cumulative noise increases proportionally. The interfering substances can also cause adduct formation promoting the spreading over multiple channels. The increase of noise when the protein size increases together with the poor peak shape are the reasons that large intact proteins are hard to analyse during top-down proteomics.

For proteomics research, the ability to search with a small mass tolerance is critical to reduce the number of theoretical possibilities and to quickly identify the correct peptide and protein [52]. Furthermore, the ability to couple it to an LC and to collect MS/MS spectra are the most critical factors for the identification of peptides and proteins. Additionally, MS/MS can be used for the identification of peptides, intact proteins or PTM’s which will be discussed in more detail down below.

(16)

16

Figure 2 Schematic of a quadrupole/Orbitrap hybrid mass spectrometer. Reprinted from [52] with permission from Wiley.

2.7. Analysis of proteins by mass spectrometry: intact proteins vs bottom-up

With the recent advancements in mass spectrometry and sample extraction techniques, LC-MS has been used increasingly for the analysis of peptides and proteins [62]. There are two major LC-MS approaches for the analysis of proteins: bottom-up peptide analysis and top-down intact analysis. During bottom-up proteomics, the protein is digested by enzymes to peptides and the peptides are analysed. This is the most common approach for protein analysis because of its high sensitivity on the commonly used triple quadrupole MS instruments. However, a downside is that it does not detect if the proteins are intact. Therefore, it can miss crucial structural information. This is especially important for the analysis of bitter peptides and the origin of these bitter peptides. The structural information of the proteins must be known to determine if the proteins are still intact before and after the preparation of the meat alternatives. Furthermore, information about post-translation modifications (PTMs) can be lost [63]. As stated in the introduction, PTMs influence bitterness. As a result of the downsides of the bottom-up approach, the top-down approach has increased in popularity.

During an intact approach, the proteins are analysed as a whole entity [61,62]. As a result, information about the structure and biotransformation are obtained. Furthermore, the sample preparation for top-down analysis can be easier because no optimisation of the digestion process is needed. However, a downside is the lower sensitivity when it is compared to bottom-up analysis. This can be a problem when starting material is limited. However, for our study, this is not a problem since soy and other legumes are easily available. Other limitations are the insufficient mass resolving power to detect differences between isoforms with minor mass differences, in particular the lower abundant ones, and limitations in software tools to facilitate protein prediction and identification [62]. Additionally, large proteins often have poor peak shape in reverse-phase LC (RP-LC). The peak shape is often improved by adding mobile phase modifiers such as trifluoroacetic acid (TFA). However, these modifiers significantly suppress the signal during ionization. An overview of the advantages and disadvantages of bottom-up and top-down proteomics is given in Table 2.

(17)

17

Table 2 Advantages and disadvantages of bottom-up and top-down proteomics

Bottom-up proteomics Top-down proteomics

Advantages Disadvantages Advantages Disadvantages

High sensitivity Does not detect intact proteins

Information about the structure and

biotransformation’s

Low sensitivity

Widely used and established

Can miss structural information No optimisation necessary of the digestion Insufficient resolving power to detect different isoforms Information about PTMs can be lost PTMs can be identified and their position

Limitations in data analysis software Large proteins are difficult to analyse

During top-down analysis, a full MS scan is done to observe the protein envelope and thus the intact mass, followed by MS/MS on one or multiple protein peaks [64]. MS/MS is done on the protein peak to fragment and thus identify the protein. There are three major method setup options for doing MS/MS on a protein with a Q-Exactive orbitrap. These options are displayed in Figure 3. In the first option, precursors are selected in a data-dependent way, leaving the decision to the instrument to pick the most abundant charge state with a narrow isolation window of 2-5 m/z (Figure 3 A). Often one defined normalized collision energy is applied to fragment the ions. A high resolution is applied to identify the fragments formed. The peaks in the MS2 need to be isotopically resolved to precisely determine their mass and thus to assign parts of the protein. The narrow isolation window is chosen to prevent the co-isolation of other proteoforms and unrelated species. This approach is mainly used for complex protein mixtures such as the soy samples used during this project.

The second approach uses a wide isolation window and stepped collision energies (Figure 3 B) [64]. When a wide isolation window of several hundred m/z is chosen, multiple charge states for one MS/MS scan can be co-isolated. The advantages of this approach is that when not all the charge states fragment equally well, the chance of including a charge state that fragments well is increased. Additionally, it increases the sensitivity by combining the common fragments from fragmenting the different charge states. Furthermore, the AGC value can be reached more efficiently, thus faster scans can be done, which is beneficial for low abundant proteins. An advantage of using stepped collision energies is that a greater variety of fragments in one MS/MS scan are obtained. Nevertheless, there are also downsides to using a wide isolation window. For example, there is a risk of co-isolating a species that does not belong to the protein. Therefore, this method is mainly used for the identification of a single protein or a low complexity sample that is chromatographically well separated.

A solution for the risk of co-isolation of different species is the third approach (Figure 3 C) [64]. In this approach, small isolation windows are used once again but multiple charge states are selected for isolation and fragmentation. Because multiple charge states are fragmented at the same time, the sensitivity is increased by combining the fragments into one spectrum. However, the overall speed of

(18)

18

the selection and fragmentation of multiple charge state is rather slow. Therefore, this method is also more suitable for deep sequencing of a low complex protein mixture. An overview of all the advantages and disadvantages of the different methods are displayed in Table 3.

Table 3 An overview of what each method can be used for and the advantages and disadvantages

Different MS Methods

Can be used for Advantages Disadvantages

Method A Complex protein mixtures Can be used for most protein samples

Basic information about the protein.

Lowest sensitivity of the different methods

Method B Single protein or low complexity samples which are chromatically well separated

Can isolate multiple charge states

Risk of co-isolating species

Increased sensitivity Not suited for complex mixes

Greater variety if fragments formed

Method C Deep sequencing of low complex mixtures

Increased sensitivity Relatively slow method

Isolates multiple charge states

During this project, intact analysis is done to confirm if the proteins are still intact in the raw materials and in the end it can also be applied to meat alternatives. This way, the identified proteins and peptides can be compared and changes can be investigated. Furthermore, it can be used to see if biologically active proteins are still intact and thus could still be active. Some of these proteins are Kunitz-type

trypsin-inhibitors and lipoxygenases. For the reasons mentioned above, method A would be the

optimal method because it is expected that the protein samples are rather complex. Therefore, a relative fast scanning speed is necessary to analyse the mixture. Furthermore, there could be overlapping proteins that would interfere with the second method.

(19)

19

Figure 3 The three different MS/MS methods for top-down analysis on a Q Exactive orbitrap. Reprinted from [64] with permission from Elsevier.

2.8. Data analysis software

Mass spectrometry instrumentation has increased in power over the last years. This increase comes with more precise and complex data. Furthermore, during proteomics research 100’s of proteins and peptides can be identified, which would be impossible to do by hand. Here comes the data analysis software in play.

There are two different ways of identifying peptides, by sequencing or by database searching. During database matching, spectra are submitted to a database where matches are selected on match scores. For sequencing, often de novo sequencing is applied. During de novo sequencing, the peptide sequence is derived from tandem MS spectra without the assistance of a database [65]. During HCD ionisation, fragmentation occurs at the peptide backbone and b- and y-ions are formed. De novo sequencing is based on calculating the mass difference between the formed fragment ions to determine the amino acid residue. One of the available software packages that combine both database searching and de

novo sequencing is PEAKS studio [66].

2.8.1. Peaks Studio [66]

During this project, PEAKS Studio 10.6 is used for the data analysis of the bottom-up method and identifying the naturally occurring peptides. PEAKS can be used for identifying peptides both digested and non-digested and for identifying proteins in a bottom-up approach. There are many other software tools available, such as MaxQuant [67]. However, there has been chosen for PEAKS because it combines both database searching and de novo sequencing and also couples the de novo sequencing

(20)

20

results to the database search results to gain more confident results. Furthermore, PEAKS is the software used at Unilever to process the data, and this way the output files and results are kept uniform.

PEAKS performs database searches by reading out FASTA files containing the proteins of interest. During this project, the glycine max (soybean) database is obtained from UniProt [68]. For database matching, the database is searched for proteins that have been digested in silico to best explain the MS/MS spectrum that is obtained [69]. Because multiple peptide spectrum matches (PSMs) may occur, PEAKS gives a score to each PSM and automatically selects the highest score. This score is determined by nine features: 1) the number of amino acids matching the de novo sequence tag; 2) the protein feature (if more peptides match to the same protein the score becomes higher); 3) the length of the peptide; 4) the sequence length per missed cleavage; 5) the sequence length per PTM in the peptide; 6) the precursor mass error; 7) the charge state; 8) the maximum length of the consecutively matched fragment ion series; 9) the number of termini that violate the enzyme’s digest rule. These features are combined with the normalized ion match score and converted to a p-value. The p-value describes the probability that a false identification in the database search achieves the same or better matching score. In short, the lower the p-value, the higher the confidence in correctly identifying the peptide. Additionally to the p-value, PEAKS uses a decoy database. This is used to estimate a false discovery rate (FDR) [69]. The conventional use of a decoy database scans a decoy database and original database separate or simultaneously. The FDR can then be made from the ratio of decoy matches and target matches. However, PEAKS concatenates the target sequence and decoy sequence and makes a new database of this. Next, it searches this new database. When it is done searching, the target and decoy identifications are separated by seeing if the match is in the first or second part of the sequence. In short, if the match is in the first part of the sequence it is a match with the target and if the match is in the second part it is a match with the decoy. The FDR is then calculated as the ratio between the number of decoy hits and the number of target hits.

For de novo sequencing it is essential to have local confidence scores because often partially correct sequences are generated due to incomplete fragmentation of the peptides [70]. PEAKS gives a local confidence score to each of the peptide residues. These scores are added up and divided by the number of residues to gain and average local confidence (ALC). The ALC gives information about the spectrum quality of the peptide.

2.7.2. Informed proteomics [71]

Informed proteomics is an open-source software package for top-down proteomics. The package consists of an LC-MS feature finder (ProMex), a database search algorithm (MSPathFinder), and a results viewer (LcMsSpectator). ProMex finds, characterises assumed proteoforms in LC-MS Data and creates features. A feature represents a group of isotopomer envelopes corresponding to the assumed proteoforms across the different charge state and LC elution times. However, individual isotope peaks often have poor peak shape when compared to the shape of the expected profile. It solves these problems by accumulating signals across different charge states and uses the LC dimension to accumulate based on retention time.

(21)

21

The detected features are then analysed by the database search tool MSPathFinder. MSPathFinder works similar to a bottom-up proteomics tool by characterizing proteoforms from the MS/MS spectra. It explores the combinatorial proteoforms space using a graph-based approach. This approach is used because proteoforms often differ by only the location of the PTMs. Therefore, the number of unique elemental composition is smaller than the number of possible proteoforms. Additionally, proteoforms have often similar fragment compositions. Last, there is a de novo sequencing algorithm implemented to find short amino acid sequences. These sequences are called tags. These tags are matched to a protein and MSPathFinder searches multiply cleaved proteoforms of the protein. This can help by reducing the search space but it may also fail to identify correct proteoforms when not enough fragment ions are detected in the MS/MS spectra.

The last part of the software package is LcMsSpectator, which is a stand-alone application that is integrated with ProMex and MSPathFinder. This software allows for interaction with both the features and the identified proteoforms.

2.7.3. ProSightPD

ProSightPD 4.0 (http://www.proteinaceous.net/) can be used within Proteome Discoverer 2.5 from Thermo Fischer Scientific [72]. ProSightPD is similar to ProSightPC, which is the benchmark for intact protein data analysis. The difference is that ProSightPC is the standalone version of ProSightPD. Within ProSighPD, a workflow for identifying intact proteins can be made. Several different workflows can be selected such as if label-free quantification has to be done or not, if the resolution of MS1 and MS2 are both high or if MS1 is low, and if an FDR has to be set. During this project, two different workflows will be used. The first one is the low/high resolution also called: “PSPD Unresolved Proteoform

Discovery Proteomics with FDR.pdAnalysis” in ProSightPD. This workflow can be used for unresolved

precursor scans and applies a 1% FDR filter. The second workflow is used when both MS1 and MS2 resolutions are high: “PSPD Comprehensive Discovery Proteomics with FDR.pdAnalysis”. This search covers all the annotated forms, unknown truncations, and unknown modifications. Again, the results are filtered to 1% FDR. Within these workflows, several settings can be modified such as the mass range, the database that has to be searched, and the deconvolution algorithm used.

After processing the samples, ProSightPD gives back multiple tables with results. The first table is the protein table. Here an overview of the different identified proteins is shown. The next table is the isoform table. If multiple isoforms are occurring from the same protein. For example, by alternative splicing or variable promoter usage [73]. The isoforms are shown here together with how many different proteoforms are detected. The next table is the proteoforms table, where proteins with truncations, containing PTMs or protein fragments are shown. The last result table is the proteoforms spectrum matches (PrSMs) table. Here the different modifications to the proteoforms are shown.

(22)

22

3. Experimental

3.1. Chemicals

Several analytical techniques have been used during this project. Down below, the chemicals are displayed per step and analytical technique. The different soy protein powders were kindly provided by Unilever.

3.1.1. Protein extraction

Sodium dodecyl sulphate (>98.5%), DL-dithiothreito l(DTT) (>99.5%), ethylenediaminetetraacetic acid (EDTA) (>99%), ammonium bicarbonate (>99.5%), urea for molecular biology, and bovine serum albumin (BSA) (≥ 96%) were obtained from Sigma. Tris-HCl has been purchased from Roche. Acetone has been obtained from VWR. ROTI®Quant Bradford solution (5x concentrated) has been purchased from Carlroth. Sodium hydroxide pellets and trichloroacetic acid (TCA) have been purchased from Merck. Last, soy protein isolate (SPI) was purchased from body&fit.

3.1.2. Denaturing size-exclusion chromatography

For the non-volatile salt buffer, sodium phosphate dibasic (≥ 99.0%) and sodium phosphate monobasic (≥ 99.5%) both purchased from sigma, were used to make a 200mM phosphate buffer at pH 6.7. For the denaturing SEC mobile phase, acetonitrile purchased from Biosolve, 0.1% trifluoroacetic acid (99%), and heptafluorobutyric acid (98%) were purchased from sigma. MilliQ water has been obtained from the Millipore system. The proteins that were used during the experiments were: bovine serum albumin (BSA) (≥ 96%), thyroglobulin (≥ 90%), γ-globulins from bovine blood (≥99%), myoglobin from equine skeletal muscle (95-100%), carbonic anhydrase from bovine erythrocytes ≥ (95%), ribonuclease A from bovine pancreas for molecular biology, HPLC peptide standard mixture, and uracil (≥99%) were all purchased from sigma. Tris-HCl has been purchased from Roche.

3.1.3. SDS-PAGE

4X LDS sample buffer, 20X Bolt MES SDS running buffer, PageRuler Unstained Protein Ladder, and Invitrogen Bolt 8% Bis-Tris plus gels were purchased from Thermo Fischer Scientific. DL-Dithiothreitol (>99.5%) and Coomassie brilliant blue G 250 for microscopy have been purchased from sigma. LC-MS grade methanol and acetic acid (glacial) (100%), have been purchased from Biosolve and Merck, respectively. MilliQ water has been obtained from the Millipore system.

3.1.4. Protein digestion for the bottom-up approach

Ammonium bicarbonate (99.5%), urea (for molecular biology), DTT (99.5%), iodoacetamide (IAA) (99%), trypsin from bovine pancreas, and TFA (99%) have been obtained from Sigma. LC-MS grade acetonitrile and water have been purchased from BioSolve.

3.1.5. LC-MS analysis

Acetonitrile (MS-grade), water (MS-grade), formic acid (MS-Grade) were obtained from BioSolve. For the assessment of the peptide analysis, a BSA (≥ 96%) digest was used. For the assessment of the intact

(23)

23

analysis, myoglobin from equine heart (≥ 90%), lysozyme from chicken egg white (≥ 90%), ribonuclease B from bovine pancreas (≥ 80%), BSA (≥ 96%), cytochrome C from equine hearth >95%, and transferrin human (≥ 98%) were purchased from sigma.

3.2. Protein extraction

Before analysis, proteins and peptides have to be extracted from the food matrix. A method has been developed to extract and precipitate proteins. After the extraction, the recovery has been determined by Bradford assay and the extraction has been controlled by SDS-PAGE [74]. The Bradford assay is a dye-binding assay based on the change of colour in response to various protein concentrations [75,76]. Coomassie brilliant blue G-250 is the dye used, and the maximum absorbance for an acidic solution shifts from 465nm to 595nm when it is bound to a protein. Therefore, a calibration curve of different concentrations of protein can be made and measured at 595nm. The more proteins are present, the higher the absorbance is. However, it is not a precise method but it gives an estimate of the protein content.

3.2.1. Extraction method

A protein extraction method based on the method of Niu et al. (2018) [77] has been developed. However, several modifications have been made. For example, PMSF is left out of the extraction buffer. PMSF is a serine protease inhibitor, which means that it prevents proteolytic degradation in general but also during the sample preparation [78]. However, for our study, it is not necessary to prevent the proteolytic degradation from proteins into peptides, because if naturally present enzymes degrade proteins before and during the sample preparation, this would also be the case during the process of making meatless products.

First, the method has been tested on soy powder from body&fit. 100 mg of SPI has been extracted with 2 mL of extraction buffer containing 1% SDS, 0.1M Tris-HCl which has been adjusted to pH 6.8, 2mM EDTA and 20 mM DTT. This has been shaken and vortexed thoroughly and kept at 4 °C for fifteen minutes. The sample was centrifuged for ten minutes at 5000 rpm. The supernatant was transferred to a fresh tube and 2mL of ice-cold 20% TCA in acetone containing 5mM DTT was added to precipitate the proteins. After precipitation, the sample was centrifuged for ten minutes at 5000 rpm. The supernatant has been removed and the precipitate has been washed with 80% acetone containing 5 mM DTT and centrifuged again. This has been repeated twice. In the end, the supernatant has been removed, the precipitate has been air-dried and dissolved in a suitable solvent for analysis.

3.2.2. Bradford assay

The protein content has been determined by the Bradford assay [74]. A BSA calibration curve from 0.1 mg/mL to 0.9 mg/mL has been made in 6M urea in 25 mM ammonium bicarbonate. For the assay, 20 µL of protein solution and 980 µL of 1x diluted ROTI®Quant Bradford solution has been pipetted into a cuvet, shaken and left for ten minutes before analysis in the spectrophotometer at 595nm. The extracted protein samples were dissolved in 6M urea in 25 mM ammonium bicarbonate and diluted accordingly to the calibration curve.

(24)

24

3.3. Size exclusion chromatography

Size exclusion will be applied to fractionate the proteins and peptides before MS analysis. When fractionating, the proteins can be separated based on size and thus different mass fractions can be made. This can be beneficial for MS analysis because peptides require a different method than intact proteins and this can increase the overall sensitivity and identification. When these proteins and peptides are fractionated in separate mass ranges, a suitable method is applied to each mass range and interference from unwanted compounds can largely be removed. Additionally, the proteins of interest have masses below 100kDa. Therefore, the larger mass fractions can be filtered out to reduce sample complexity. The proteins of interest are mainly: glycinin, beta-conglycinin, 7S globulin,

albumins, trypsin-inhibitors, and lipoxygenases. Lipoxygenases are specifically interesting because they

can cause changes in flavour and aroma. Additionally, large proteins are difficult to analyse by MS, which would induce more background noise.

3.3.1. Column and method

A Waters Acquity UPLC consisting of a sample manager, binary solvent manager, TUV detector, and a Waters fraction collector III has been used for the SEC separation and fractionation on a TSKgel SWXL 7

µm, 6 mm internal diameter x 40 mm length guard column and a TSKgel G3000SWXL 5µm, 7.8mm

internal diameter x 300 mm length (250Å) analytical column.

The salt-based SEC method was run with a 200 mM phosphate buffer at a flow of 0.8 mL/min and 0.5 mL/min. Proteins were detected at 214nm. The denaturing SEC method is based on the methods of Vandenheede et al. (2019) [50] and Liu et al. (2009) [51]. The proteins have been separated based on size with 20% acetonitrile and 0.1% TFA with a flow of 0.5 mL/min and proteins have been detected at 214 nm. For optimisation and calculating the calibration curve, 10 µg of protein standard has been injected on the column and 1 µg of uracil. For the separation of soy protein samples, 100 µg sample has been loaded onto the column. The standards have been solved in mobile phase. The protein samples have been solved in mobile phase when the samples were non-reduced.

When reducing the proteins before fractionation, the samples were solved in 10mM of Tris-HCl at pH 7.5. After solving, DTT was added to reduce the proteins and the sample was heated at 95 °C for 10 minutes. Before fractionation, similar amounts of mobile phase was added to double the volume of the sample. The final concentration of the sample was 2 mg/mL.

3.3.2. Fractionation of the proteins

The different protein samples have been fractionated by attaching a Waters fraction collector III after the TUV detector. Soy powder 4 and 6 have been fractionated to asses with SDS-PAGE. 100 µg of soy protein sample has been loaded on the column and has been fractionated over 9 fractions with 2 minutes per fraction. Fractions 2 through 8 have been freeze dried and solubilized in a suitable SDS-PAGE solvent. Besides collecting fractions for SDS-SDS-PAGE, fractions of 100 µg soy powder nr. 3 and 100 µg soy powder nr. 6 have been collected for intact protein analysis and bottom-up proteomics.

(25)

25

In total, six fractions, with the first fraction being two minutes and the other fractions being four minutes, starting at twelve minutes have been collected. This has been repeated 6 times. 4 mL of fractions 1, 2, and 3 of the in total 12 mL collected has been freeze-dried and used for bottom-up proteomics. The remaining 8 mL has also been freeze-dried and was used for intact protein analysis. This has only be done for fractions 1, 2 and 3 because fractions 4, 5, and 6 should only contain naturally-occurring peptides and thus bottom-up proteomics would not be necessary to identify the peptides. The first three fractions have also been collected to analyse with SDS-PAGE.

3.4. Sodium dodecyl sulphate polyacrylamide gel electrophoresis

SDS-PAGE has been applied to assess the extraction and fractionation. For the extraction, SDS-PAGE has been done to compare the extracted SPI proteins with the untreated SPI. For the fractionation, SDS-PAGE has been done to assess if the fractionation works properly and thus a decrease in mass can be seen for the different fractions. Furthermore, it has been applied to asses the fractionation of large proteins from the soy protein powders.

3.4.1. The SDS-PAGE method used

The SDS-PAGE has been run as suggested by Thermo Fischer Scientific [79]. First, the MES SDS running buffer and the LDS sample buffer were diluted to a 1X solution with MilliQ water. 50mM DTT has been added to the LDS sample buffer as a reducing agent. The samples were dissolved in the LDS sample buffer, heated to 70 °C and shaken at 300 rpm for ten minutes. After heating, the samples were centrifuged for ten minutes at 10,000 rpm. The samples were stored overnight at 4°C. The next day, the gel ran for 20 minutes at 200V. For the protein ladder, 5 µL or 7 µL has been pipetted onto the gel. For the samples, between 50 µg and 100 µg of protein sample has been loaded onto the gel. After 20 minutes, the case was removed and the gel was washed shortly with MilliQ water followed with three times a five-minute wash with 20% methanol and 10% acetic acid. After washing, the gel was stained for 1.5 hours with a Coomassie blue staining solution consisting of 40% methanol, 10% acetic acid, and 0.1% Coomassie brilliant blue G 250. After staining, the gel was rinsed three times with water. Last, the gel was destained three times with 20% methanol and 10% acetic acid and the final wash was left overnight.

3.5. Identification of soy proteins and peptides

Down below, the methods used for the identification of naturally occurring peptides, digested peptides from the bottom-up approach, and the intact proteins during the top-down analysis will be described. This includes the bottom-up method, the LC-MS/MS analysis of both the peptides and intact proteins, as well as the settings used in the data processing software. PEAKS Studio [66,69,80] is used for the identification of naturally occurring peptides and digested peptides. ProSightPD 4.0 within Proteome Discoverer 2.5 from Thermo Fischer [72] is used for the processing of the top-down analysis.

(26)

26

3.5.1. Protein digestion

The method used for protein digestion is a method supplied by the University of Amsterdam. First, a 25mM ammonium bicarbonate solution is made. Everything is solved in this solution unless stated otherwise. Next, the proteins, which can be a BSA standard or the protein fractions collected, were solved in a 6M urea solution. The protein fractions were solved in 100 µL of 6M urea. After solving the proteins, 5 µL of a 200mM DTT solution was added and the Eppendorf was left at 37 °C for an hour. After an hour, 20 µL of a 200mM IAA solution was added. The Eppendorf was kept away from light and left for an hour at 37 °C. Final, 20 µL of the DTT solution was added together with 900 µL of the bicarbonate solution and 13.6 µL of a 0.1 mg/mL trypsin solution. This was left overnight at 37 °C. The next day, solid-phase extraction (SPE) has been applied to desalt the samples. C-18 Discovery 1 mL SPE columns were conditioned with 3 times 1 mL acetonitrile and washed with 3 times 1 mL 0.1% TFA in water. The sample was loaded 500 µL a time and washed with 3 times 1 mL 0.1% TFA solution. The analyte was eluted with 2 times 200 µL 50% ACN with 0.1% TFA, and 3 times 200 µL of 75% ACN with 0.1% TFA. After the SPE, the samples were freeze-dried overnight and the samples were stored at -20 °C until analysis.

3.5.2. LC-MS/MS analysis of peptides

The natural occurring peptides and digested proteins have been separated on a Thermo nano-LC Q Exactive-MS. After freeze-drying, the peptides have been reconstituted in 100 µL of 2% ACN and 0.1% TFA. The mobile phases consist of A: 2% ACN and 0.1% FA, B: 80% ACN and 0.1% FA, and for the loading mobile phase: 2% ACN with 0.1% TFA. First, 20 µL of sample has been loaded on the trap column with a flow of 10 µL/min and focussed for 5 minutes. After loading, the samples have been separated on a C18 nano-LC column (75 µm x 150 mm) at 45°C with a flow of 0.300 µL/min and a 120-minute gradient. The initial concentration of mobile phase B of the gradient was 2% and this increased to 50% in 90 minutes, followed by a steep increase to 80% at 98 minutes and this was held for one minute. The column was quickly washed by decreasing to 2% B in one minute and increasing to 60% B again in one minute. Last, the column was equilibrated at 2% B for 18 minutes.

The following MS method is used to detect natural peptides. It was obtained from and optimized at Unilever and is adapted to combine with nano-LC. The parameters are: for MS1, the scan range was 100-1500 m/z, max. injection time of 60 ms, an AGC target of 36_{and a resolution of 70k. For MS2, the}

scan range starts at 100m/z, max. injection time of 100 ms, an AGC target of 25_{, a resolution of 17.5k,}

and stepped collision energies of 15, 30, and 45. The charges that were excluded are unassigned, 7, 8, and >8. The minimum AGC target for the dd-setting was 83_{. The following MS tune settings were used:}

capillary temperature of 250 °C, max spray current of 50V, spray voltage of 2100V and an S-lens RF level of 50. For the bottom-up method, the only difference was that more charges were excluded, specifically: 1, 7, 8, and >8

3.5.3. Peaks Studio

First, the bottom-up approach. The first step in the workflow is data refinement. Here the option for “correct precursor” is selected, with the masses 2 through 8 selected, to reduce the amount of data

(27)

27

processing needed. The next step in the workflow is the search algorithm. A precursor mass tolerance of 15 ppm has been selected and a fragment ion tolerance of 0.05 Da. Trypsin has been selected as the enzyme used and the samples have been searched against the Uniprot Soybean database [81]. A maximum of three missed cleavages and five variable PTMs per peptide were allowed. The false discovery rate (FDR) for the peptides and proteins has been set to 1%. For the identification of proteins, two unique peptides had to be present. This is to filter out random matches with similar proteins. The

de novo only score has been set to ≥70%.

For the naturally occurring peptides, a very similar workflow is applied. However, instead of selecting trypsin as an enzyme, no enzyme has been selected. Therefore, the digest mode was set to unspecific and the max missed cleavages was automatically set to 100. All the other settings were kept the same.

3.5.4. Intact protein analysis

MS-Method optimisation

The MS-method has been optimised using a protein mixture containing five proteins: Transferrin,

BSA, myoglobin, lysozyme, and Ribonuclease B. First, the influence of the protein mode and changing

the trapping gas pressure has been examined. The trapping gas has been assessed with cytochrome C as a standard. From here, the following parameters have been optimised: SID, resolution in both MS1 and MS2, the collision energy, and the injection time. All these parameters have been changed one by one and the most optimal settings were used for the analysis of the proteins. The final method is described below and the exact values that have been optimised are shown in Table 4.

Table 4 Overview of all the optimised settings together with the most optimal value.

Setting Values tested Most optimal value

Trapping gas pressure 0.2, 0.5 and, 0.8 0.2

SID 10, 15, and 20 15

Resolution MS1 17.5k, 70k, and 140k 17.5k for protein mass above 40 kDa. 70k

or 140k for the masses below 40kDa

Resolution MS2 70k and 140k 70k. Isotopically resolved MS2 is needed

for the software to assign MS/MS spectra

NCE Fixed: 10, 15, 20, and

different stepped collision energies

A collision energy of 20 is optimal. Most suitable for a wide range of proteins. A fixed collision energy is also suggested in the literature [64].

Injection time MS 1 50ms and 100ms 50 was often enough to reach the AGC

target.

Injection time MS 2 250ms, and 800ms 800 ms seemed optimal between having

a high enough scan speed and collecting enough ions when the AGC target was not reached.