Increased solubility and functional properties of precipitated Alfalfa protein concentrate subjected to pH shift processes

Protein fractionated from alfalfa exhibits promising properties for exploitation as food ingredient. However, further knowledge and improvement of the functional properties is essential to increase the potential for commercial use. This might be obtained by acid or alkaline pH shift processes. The effect of pH shifting compared to controls (buffer solutions) at different pH-values were investigated concerning protein solubility, foamability, and gelation. The effect of salt (NaCl) concentration was also studied. At pH 12, more cross-linkages were observed as lysinoalanine, lanthionine and also conversion of L-Ser to D-Ser was highest at pH 12. Therefore, the pH-shift method with pH 11 was chosen for further investigation, which increased the solubility from 50% to 70 – 80%. The alkaline pH shift (pH 11) increased the foam overrun and stability, where pH 7 reached the highest stability with > 60% foam volume remaining after 1 h. The protein gel strength showed remarkable potential with the alkaline pH shift method with re-adjusting to pH 7, reaching 2584 Pa with 72 g/L protein. For the acid pH shift with re-adjustment to pH 2 followed by pH 3, the foam overrun increased to 140% and protein solubility to 57% at low protein concentration (2 g/L). Addition of salt (185 mM NaCl) increased the foam overrun at pH 5 and 7 for all samples. Overall, these results indicated that prior alkaline pH treatments and salt can increase the potential of alfalfa protein in terms of certain functionality in foods.


Introduction
Due to the increasing demand for food along with a desire for sustainable food options, several plants have been investigated as potential protein sources.Protein from seeds such as pea and soy have had the most progress and are growing on the market (Day, 2013).However, some plants such as red clover and alfalfa can reach higher protein yields per hectare and contain all essential amino acids, whereas seeds often are insufficient in amino acids containing sulfur (Bals, Dale, & Balan, 2012;Dijkstra, Linnemann, & Van Boekel, 2003).These plants contain the highly nutritious protein rubisco (ribulose 1,5-bisphosphate carboxylase), which is essential for photosynthesis and is the most abundant protein in nature (Ellis, 1979;Raven, 2013).Rubisco is yet an underutilized protein in regard to food, which is mostly due to association of the protein with compounds that are unwanted in a food context such as polyphenols and chlorophyll in the leaves.Thus, processing of the plant is essential for human consumption, which may alter the nutrient quality and food functionalities (Di Stefano, Agyei, Njoku, & Udenigwe, 2018).Alfalfa, a perennial plant, has been widely studied to create a fraction valuable as food ingredient (Di Stefano et al., 2018;Knuckles, Bickoff, & Kohler, 1972;Lamsal, Koegel, & Boettcher, 2003).To be acceptable as food ingredient, nutritional quality and taste are of foremost importance followed by desirable functional properties.Functionality of protein fractionated from alfalfa has been evaluated in different studies including solubility, foamability, emulsification, gelation, water holding and flavor binding.Of those properties, the solubility is one of the most important parameters.Highest solubility on alfalfa extracts has been observed at pH above 7 or below 3 reaching from 40 to 90% protein depending on the method of extraction (Hojilla-Evangelista, Selling, Hatfield, & Digman, 2017; Knuckles & Kohler, 1982;Lamsal, Koegel, & Gunasekaran, 2007).Heat coagulated and acid precipitated extracts show the lowest solubility (Baraniak, 1990;Lamsal et al., 2007), whereas ultrafiltrated concentrates have increased solubility, possibly due to a more retained native protein structure (Lamsal et al., 2007).In aerated foods in the categories of bakery, confectionary and beverages, the foaming of proteins contributes to the structure.Foaming stability of foams prepared using alfalfa extracts has been observed to be highest at the isoelectric point around pH 4.5.This is due to less electrostatic repulsion resulting in increased interactions between proteins and thus stronger films surrounding the air bubbles (Knuckles & Kohler, 1982;Lamsal et al., 2007).An improved foam stability compared to that of egg white, whey and soy protein has been reported around this pH, whereas soy and whey protein are superior to that of alfalfa extracts at pH 7 (Knuckles & Kohler, 1982;Martin, Castellani, de Jong, Bovetto, & Schmitt, 2019).Gelation of proteins is of great importance in many food matrices such as dairy products, bakery, and meat products as it creates the structural basis for holding water, sugars and flavors (Wang & Kinsella, 1976).Alfalfa protein extracts exhibit great potential for gelation due to the low denaturation temperature of 72.6 • C (Tomimatsu, 1980) of rubisco and a lower critical concentration for gel formation compared to that of soy and whey protein (Martin, Nieuwland, & De Jong, 2014).However, this only applies to neutral pH as both lower and higher pH result in a weaker gel (Knuckles & Kohler, 1982;Lamsal, Koegel, & Gunasekaran, 2005).A low denaturation temperature can be an advantage in relation to many food matrices as bakery and meat products, where foaming and gelation is involved as these properties are improved upon partial denaturation.However, it can also constitute a challenge in for instance beverages that need thermal processing, as this may cause precipitation of the protein and affect the sensory quality (Damodaran, 2017).Hence, the low denaturation temperature might limit the utilization in some products.
Although, alfalfa protein seems promising as a novel protein source, further knowledge of its properties at different pH conditions is essential for valorization.Subjecting proteins to acid or alkaline pH shift has been observed to result in increased functionalities for egg albumen and soy protein isolate due to altering of the structural conformations (Jiang, Chen, & Xiong, 2009;Jiang, Xiong, & Chen, 2010;Liang & Kristinsson, 2007).Alkaline and acid pH shift refers to adjusting the pH to extreme acidic or alkaline pH followed by readjustment to neutral pH.This causes increased repulsions between side chains resulting in partial unfolding of the tertiary structure of proteins, which is referred to as the "molten globule" structure (Kristinsson & Hultin, 2003).However, subjecting proteins to pH shift processes can also result in racemization and cross-linking, which can decrease the nutritional value and the digestibility (Gilani, Xiao, & Cockell, 2012;Schwass & Finley, 1984).
The effect of pH shift on the functional properties of alfalfa protein has not yet been evaluated.The objective of this study was to investigate alkaline and acid pH shift processes on alfalfa protein concentrate in relation to different functionalities, compared to the protein prepared in buffer solutions (controls) at different pH-values.Furthermore, the effect of different concentration of salt (NaCl) was studied, which up to now have only been briefly touched upon in relation to emulsification, solubility and gelation of alfalfa protein (Knuckles & Kohler, 1982;Wang & Kinsella, 1976).Salts are often used as co-ingredients in food matrices for both food safety, sensory and functional reasons, and thus, knowledge of the interactions between protein and NaCl at different pH-values is essential for food applications and a valuable step towards commercial implementation.

Plant material and protein extraction
Fresh alfalfa (Medicago sativa) was harvested in the middle of August 2019 using a MaksiGrass (MaksiGrass, Skjern, Denmark).The whole plants were fractionated at the same day in a screw press (Angel Juicer Mini Commercial 20 K, Domotech, Denmark).The plant juice was cleared by removing fibers and debris by centrifugation (10 min at 1700g, 4 • C).Citric acid (2 M) was added dropwise to the cleared juice to a final pH of 4.5 (pI of rubisco) and samples precipitated during storage at 4 • C overnight.The next day centrifugation (15 min at 1700g, 4 • C) was used to isolate the protein pellet, which was frozen and freeze-dried, resulting in a green powder.The freeze-dried alfalfa protein powder was stored at − 18 • C.

Protein determination
Total nitrogen content of freeze dried alfalfa protein powder was analyzed in duplicates by Dumas combustion (DUMATHERM®, Gerhart Analytical Systems, Königswinter, Germany) using 1.6 mg O 2 /mg sample and an O 2 flowrate of 200 mL/min.The protein content was calculated using a nitrogen factor of 6.25.

Optimization of alkaline pH shift
The alkaline pH shift was performed as described by Jiang et al. (2009) with slight modifications.Briefly, suspensions with a final protein content of 10 g/L were prepared in Milli-Q water and the pH adjusted to 7, 8, 8.5, 9, 10, 11 and 12, respectively with 1 and 2 M NaOH.The suspensions were stirred for 2 h at room temperature followed by adjusting the pH to 7 with 1 and 2 M HCl and 30 min of stirring.Solutions were used for protein solubility, L/D analysis and process-induced changes on amino acids to evaluate the best performer.

L/D-amino acids analysis
Solutions from 2.4 were centrifuged at 4000g for 20 min at 4 • C in duplicates and the supernatant was freeze dried to investigate the soluble protein.The relative ratio of L-and D-amino acids of the soluble protein was determined by liquid chromatography mass spectrometry (LC-MS) analysis according to Danielsen, Nebel, and Dalsgaard (2020).In brief, 25 mg protein powder was hydrolyzed into amino acids using DCl (deuterated hydrochloric acid) in 1 mL vacuum hydrolysis tubes (Thermo Scientific, IL, USA).A chiral derivatization with S-NIFE of the hydrolyzed samples allowed a subsequent separation of the L-and D enantiomers on reverse phase HPLC.LC-MS/MS analysis was performed on a triple quadrupole tandem mass spectrometer (6460 TripleQuad LC/MS, Agilent Technologies, Santa Clara, CA, USA) coupled to a 1290 Infinity LC system (Agilent Technologies, Santa Clara, CA, USA).Chromatographic separation was carried out on a Luna Omega C18 column (100 × 2.1 mm, 1.6 μm, 100 Å) (Phenomenex, Torrance, CA, USA).The relative ratio of D-enantiomers was calculated based on the integration of the L-and D enantiomer peaks in the same MS spectrum.

Process-induced changes on amino acids
Soluble protein obtained in 2.5 were also analyzed for process induced changes on amino acids (Carboxymethylysine (CML), carboxyethyllysine (CEL), furosine (FUR), lysinoalanine (LAL) and lanthionine (LAN)).An amount of 50 mg in duplicates was hydrolyzed using 1 mL of 1% (V/V) mercaptoethanol and 3% (W/V) phenol in 6 mol/L HCl in vacuum hydrolysis tubes (Thermo Scientific prod.29570) at 110 • C for 20 h.After hydrolysis the samples were cooled at 4 • C and transferred to an Eppendorf tube and centrifuged for 5 min at 20800g (Eppendorf 5417 R centrifuge).The samples (500 μl) were neutralized with 350 μl 6 mol/L NaOH, filtered through a Whatmann Mini-UniPrep 0.2 filter vials and analyzed with LC-MS.
The quantitative analysis was performed on an 8050 triple quadrupole (QQQ) mass spectrometer (Shimadzu, Kyoto, Japan) coupled to a Nexera X2 LC system (Shimadzu, Kyoto, Japan).The LC was equipped with an Intrada, Amino Acid, 150 × 3 mm, Imtakt (Prod.WAA35) purchased from Biolab DK.Solvent A was 0.1% formic acid in acetonitrile.Solvent B was 100 mmol/L ammonium formate.The compounds were eluted from the column using a flow rate of 0.6 mL/min with a linear gradient 17% B at 0 min to 100% B at 16 min.To avoid carry over, solvent B was at 100% until 23 min after which 17% B was reached at 25 min.The column was equilibrated for 10 min (35min) at 17% B before next injection.The oven temperature was 35 • C.
The electrospray ionization was set as followed; heating gas flow 10 mL/min, interface 300 • C, DL 125 • C, heat block 250 • C, and drying gas flow 10 L/min.The precursor ions were identified running standards and the fragment ions were automatically generated in the LCMS solution 5.97 SP1 software.The MRM transitions are presented in Table 1.The quantification was obtained from external calibration curves (10-2500 ng/mL for CML, CEL, and LAN, 100-25000 ng/mL LAL, and 1-250 ng/mL for FUR) containing the same IS mix as described for the sample preparation.CEL D 4 IS was used for CEL and LAN, CML D 2 IS was used for CEL, FUR D 4 IS was used for FUR and LAL.

Preparation of protein solutions
Protein solutions were prepared at pH-values of 3, 5 and 7 both in buffer solutions (controls) and after alkaline pH shift (pH 11) with a protein content of 2 and 10 g/L and with and without addition of salt (Table 2).For the samples without salt, buffers were prepared with an ionic strength of 15 mM for the samples with a protein content of 2 g/L and 50 mM for the samples with 10 g/L as this was equal to the ionic strength resulting from the alkaline pH shift at the respective protein concentrations.For the samples with salt, NaCl was added to reach a concentration of 185 and 150 mM for 2 and 10 g/L protein, respectively, resulting in a final ionic strength of 200 mM.Furthermore, an acid pH shift was performed referred to as pH 2 → 3. The acid pH shift was only performed at pH 2 → 3 due to precipitation when the pH was raised above pH 3. The different combinations of solutions showed in Table are described in the following subsections (2.7.1-2.7.3).

Solubilization of protein in buffer
The controls were made in different buffer systems; 30/100 mM trisodium citrate dihydrate/citric acid (pH 3), 22/75 mM sodium acetate/acetic acid (pH 5) and 7.5/25 mM disodium hydrogen phosphate dehydrate/sodium dihydrogen phosphate (pH 7).These buffers were prepared with concentration reaching an ionic strength of 15 mM for the samples with a protein content of 2 g/L and 50 mM for the samples with 10 g/L.Furthermore, solutions were made with and without the addition of NaCl, where 185 and 150 mM NaCl were added to 2 and 10 g/L protein solutions, respectively.Controls at different pH were stirred for 2 h and pH was adjusted if necessary.

Acid pH shift
For the acid pH shift, protein was solubilized in water giving a final concentration of 2 or 10 g/L.NaCl was added in a concentration of either 15 or 50 mM for 2 and 10 g/L, respectively, to achieve a similar ionic strength as protein prepared in buffer or with alkaline pH shift.The pH was adjusted to pH 2 with 1 M HCl and stirred for 1 h before readjustment to pH 3 with 2 M NaOH and stirred for 1 h.Additionally, solutions were made with further addition of NaCl, where 185 and 150 mM NaCl were added to 2 and 10 g/L protein solutions, respectively.

Alkaline pH shift
Samples with alkaline pH shift were prepared in Milli-Q water with a final protein concentration of 2 or 10 g/L.The pH was adjusted to with 1 and 2 M NaOH and stirred for 2 h at room temperature followed by adjusting the pH to 2 → 3, 3, 5 and 7 with 1 and 2 M HCl and 30 min of stirring.For the pH 2 → 3, adjusting of pH was performed as in 2.7.2 after reaching pH 11.Furthermore, solutions were made with and without the addition of NaCl, where 185 and 150 mM NaCl were added to 2 and 10 g/L protein solutions, respectively.
The effect of ionic strength on solubility was evaluated for the alkaline pH shift process with pH 11 and readjustment to 7. NaCl was added to reach a final concentration of 50, 100 and 150 mM.This was performed in two ways; addition of NaCl before and after alkaline pH shift to suspensions with a protein content of 10 g/L.

Conductivity
The conductivity of the different protein solutions described in 2.7 was measured using a CDM210 Conductivity Meter (Radiometer, Copenhagen, Denmark).Each solution was shaked well before measuring and samples were measured at 20 • C in duplicates.

Protein solubility
Protein solubility of the different protein solutions described in 2.4 and 2.7 was evaluated.All suspensions were centrifuged at 4000g for min at 4 • C and prepared in duplicates.Protein content of the supernatants was measured in duplicates by a modified Lowry procedure following the technique of Markwell, Haas, Bieber, and Tolbert (1978).Sample or standard (BSA) was diluted to 10-100 μg protein/mL in 0.1 M NaOH.Hereof, 200 μL of dilution was mixed with 600 μL solution containing 1 part 4% CuSO 4 ⋅5H 2 O mixed with 100 parts 2.0% Na 2 CO 3 , 0.4% NaOH, 0.16% sodium tartrate, 1% SDS and incubated at room temperature for 30 min.Folin-Ciocalteu phenol reagent was diluted 1:1 with Milli-Q water and 60 μL was added to each sample and mixed well followed by incubation for 45 min at room temperature.
From this, 250 μL was transferred to a microwell plate in triplicates and read at 750 nm with a Cary 60 UV-Vis spectrophotometer (Agilent technologies).The solubility was calculated as percent of protein present in the supernatant.

Protein composition by SDS-PAGE
The protein composition of protein solutions (control and alkaline pH shift) with 2 g/L protein without addition of salt at pH 2 → 3, 3, 5 and 7 were evaluated by SDS-PAGE following the technique of Laemmli (1970) under reducing conditions.Protein samples were diluted 1:1 with SDS-PAGE sample buffer and heated at 90 • C for 2 min and run on Criterion TGX Stain-Free Precast Gel 4-15% (Bio-Rad, Hercules, USA).The gel was fixed in 50% ethanol, 8% phosphoric acid and stained with colloidal Coomassie Brilliant Blue following the principle of Kang, Gho, Suh, and Kang (2002).

Zeta-potential
The zeta-potential (mV) of protein solutions (control and alkaline pH shift) with 2 g/L protein with (185 mM NaCl) and without addition of salt at pH 2 → 3, 3, 5 and 7 were measured by a Stabino® system (Particle Metrix, Meerbusch, Germany).Suspensions were left for 30 min to sediment coarse undissolved particles before measurements.
Measurements were performed for 120 s with a 200 μm piston and the endpoints were recorded.Four measurements were completed per sample.

Foaming properties
Foaming properties were evaluated for protein suspensions (control and alkaline pH shift) with 2 g/L protein with (185 mM NaCl) and without addition of salt at pH 2 → 3, 3, 5 and 7.The procedure followed the technique of Hammershøj, Peters, and Andersen (2004), where a 20 mL suspension was shaked in a 100 mL closed graduated cylinder for 45 s at a frequency of 4 Hz.Foamability was evaluated as Foam Overrun (FO) as calculated by equation ( 1), based on the volume of foam after min (V foam ) and the initial volume of the suspension (V liquid ).The volumes of foam and liquid were subsequently observed for 1 h and the foam stability was calculated as relative foam volume (FV) remaining after 1 h, equation (2).Samples were run in duplicates. (2)

Gelation monitored by oscillation
Protein suspensions with a concentration of 36 g/L were prepared as controls and with alkaline pH shift at pH 2 → 3, 3, 5 and 7. Sample solubilization for the controls was done in both 15 mM buffer and with the addition of 185 mM NaCl.The pH was checked and adjusted with M NaOH and 1 M HCl.For the alkaline pH shift, sample solubilization was done in 200 mM NaOH to obtain the target pH of 11.For pH 5 and 7, protein suspensions were also studied with a protein concentration of 72 g/L, where 400 mM NaOH was needed for the alkaline pH shift to obtain the target pH of 11.
The gelation process was evaluated by small deformation rheology with a frequency of 0.5 Hz and 0.2% strain following the procedure of Schmidt et al. (2019) with slight modifications.The deformation parameters of frequency and strain were settled to be within the linear viscoelastic region (LVR) based on initial tests of the alfalfa protein solution before and after gelation (data not shown).For the measurements, a rheometer (AR G2, TA instruments, New Castle, DE, USA) equipped with a cup-and-bob geometry (cup radius 15 mm, bob radius 14 mm) was used.A volume of 20 mL of sample was transferred to the geometry and allowed to equilibrate for 2 min at 20 • C, before heating to 85 • C (3 • C/min), holding for 20 min at 85 • C followed by cooling to 20 • C (3 • C/min) and holding at 20 • C for 20 min.The storage (G ′ ) and loss (G ′′ ) moduli were recorded during the total run time.Samples were run in duplicates.

Statistical analysis
Results are presented as means with standard deviations.One-way and multi-way analysis of variance (ANOVA) with different parameters depending on the specific experiment (solution media, pH, salt concentration, protein concentration) were carried out.Differences were regarded as significant at a minimum level at 95% (P < 0.05) A Duncan test was used to create a compact letter display (CLD) to designate significantly different averages.Statistical analysis was carried out using RStudio (RStudio Team, 2020).

Optimization of alkaline pH shift
From 1 kg of fresh plant material an amount of 145 g wet protein paste was obtained, which was further freeze-dried into 19.9 g dry protein extract.The freeze-dried extract had a crude protein concent of 42.2 ± 0.2%, which aligns with previous literature using acid precipitation (Baraniak, 1990; Stødkilde, Damborg, Jorgensen, Laerke, & Jensen, 2019).The protein extract was subjected to different conditions (pH shift/control, pH, salt, protein concentration) to test the solubility, which relies on the extent of protein-water interactions compared with protein-protein interactions.Fig. 1 illustrates the effect of alkaline pH shift on protein solubility, when the solution was raised to different pH before readjustment to pH 7. For alkaline pH shift with pH 7 to 9, the protein solubility was 46-48%, whereas pH 10, 11 and 12 resulted in 53%, 68% and 90% protein solubility, respectively.This increase in solubility can be due to partial hydrolysis caused by the alkaline environment, which as well can lead to racemization and cross-linking of the proteins (Betschart, 1974;Schwass & Finley, 1984).Thus, the ratio of Land D-amino acids and process-induced changes on amino acids were investigated (Figs. 2 and 3).
The relative ratio of L-and D-amino acids was measured for 10 different amino acids (Fig. 2).The treatment (pH shift process or protein extract) had a significant impact (P < 0.001) on the level of D-amino acid for some of the amino acids.Furthermore, the interaction between treatment and amino acids had significant impact (P < 0.001).For alanine, an increasing pH resulted in decreasing levels of D-amino acids (P < 0.001), whereas the extract reached the minimum value of D-amino acid.For arginine, histidine, lysine and valine, an increasing pH resulted in increasing levels of D-amino acids (P < 0.001), while the extract reached values comparable to that of pH 11.This could be due to an amount of D-isomer of threonine in the extract that was not soluble and hence does not appear in the pH shift processes.Increasing levels for lysine and valine with increasing pH have previously been observed for alfalfa protein (Schwass & Finley, 1984).For proline and threonine, the levels of D-amino acids showed no significant differences on pH-level, which aligns with previously literature (Schwass & Finley, 1984).Tyrosine had a smaller impact (P = 0.0471) with pH 12 and the extract reaching the maximum values.The amino acids methionine and serine showed a significant (P < 0.001) maximum D-amino acid level at pH 12 compared to the other treatments, with serine presenting the highest level of ~1.3%.Previous literature (Schwass & Finley, 1984) has observed that especially serine in alfalfa protein concentrate is susceptible to racemization upon alkaline conditions, which explains the higher levels found at pH 12 in this study.Methionine has also previously shown to undergo racemization (Jasour et al., 2017).For most of the amino acids, a raise of pH in the alkaline pH shift resulted in higher levels of D-amino acids as anticipated.However, for alanine the opposite occurred contradicting the results of Schwass and Finley (1984) and further investigation is needed to understand the underlying mechanism.The levels of D-amino acid reported in Schwass and Finley (1984) reached 40% at pH 12 for some of the amino acids, whereas in this study the highest D-amino acid content was 1.3% for serine at pH 12.This is due to the differences in methods analyzing the content of D amino acids.When analyzing a sample for D-amino acids, a hydrolysis is needed.Schwass and Finley (1984) uses a regularly acidic hydrolysis, which results in overestimation of D-amino acids due to racemization of the sample during hydrolysis.In this study, the method of Danielsen et al. (2020) was applied, where deuterium chloride (DCl) was used for the hydrolysis.Thus, those amino acids that underwent racemization during the acid hydrolysis increased in mass by +1 Da and was consequently not included in the analysis.This gave a more accurate determination of the L-and D-amino acid ratio and thus these values cannot be directly compared with the values in Schwass and Finley (1984), where they did not count for artificial amino acid racemization during acid hydrolysis, but only be compared in terms of tendencies.
Process-induced changes on amino acids were analyzed by measuring levels of lanthionine (LAN), lysinoalanine (LAL), carboxymethyllysine (CML), carboxyethyllysine (CEL) and furosine.LAN and LAL are cross-linked amino acids that can be formed at high temperatures or at high pH.This is due to racemization of amino acids resulting in the highly reactive intermediate dehydroalanine (DHA), which can further react with either lysine forming LAL or cysteine forming LAN (Damodaran, 2017).For both LAN and LAL, a significantly higher content was observed at pH 12 compared to alkaline pH shift at lower pH and in the protein extract, especially for LAL reaching ~600 ng/mg protein (Fig. 3).Schwass and Finley (1984) detected a small content of LAL in alfalfa extract subjected to pH 8.5 and also reported an increase for pH 10 and 12.8, but in higher levels (~4000 ng/mg protein) than observed in this study.CML, CEL and furosine are amino acid derivatives that in general are related to oxidation and heat treatment.However, these also seemed to be affected by the alkaline environment as higher values were observed for both pH 11 and 12.As this analysis was performed on the soluble protein, one must expect a higher content of cross-linked amino acids when the precipitate is included.Schwass and Finley (1984) investigated the entire protein solution, which may explain the higher levels of LAL reported in their study.Based on these results, a pH of 11 was chosen for the alkaline pH shift as this increased the solubility but did not result in major cross-linking and amino acid racemization as for pH 12.

Protein solubility
The effect of salt concentration was investigated by two concepts; addition of salt before and after the alkaline pH shift at pH 11.Salt decreased the solubility (no salt vs. addition of salt), but higher protein solubility was observed when the salt was added after the alkaline pH shift compared with before (Fig. 4).Addition of low amount of salts (<0.5 M) can affect the solubility in two different ways depending on the protein.If the protein contains a high amount of nonpolar patches, the solubility will decrease due to enhanced hydrophobic interactions, whereas it increases for those that do not due to higher repulsive electrostatic interactions (Damodaran, 2017).Thus, results indicate that this alfalfa protein extract contains a high amount of nonpolar patches as previously documented by Barbeau (1990).Furthermore, the solubility decreased significantly with increasing addition of salt before alkaline pH shift (55%-49%), whereas samples with addition of salt after alkaline pH shift showed no significant difference in solubility with increasing addition of salt.This might be caused by interactions between the salt and the molten globule state of the protein, thus exposing more hydrophobic patches that may further interact at high ionic strength conditions.However, this hypothesis needs additional investigation to understand the mechanism.
A range of pH values (pH 2 → 3, 3, 5, 7) was tested with and without alkaline pH shift procedure to gain further knowledge of pH values relevant for different food systems (Fig. 5).Results showed significantly higher solubility for the samples subjected to alkaline pH shift and readjusted to pH 7 without salt at a protein content of 2 g/L (80%), compared to the control at pH 7 (45-50%).For the higher protein content (10 g/L), a lower solubility of 70% was found at pH 7 after alkaline pH shift without salt.The samples with addition of salt at pH 7 showed lower solubility (Fig. 5A), whereas at higher protein concentration (Fig. 5B), the control reached similar values with and without salt (50%).Solubility at pH 5 reached 30-40% for both controls and alkaline pH shifts, whereas a significantly higher solubility was observed for the samples with addition of salt possibly due to the principle of salting in.Higher solubility was also observed with the lower protein content of 2 g/L.At pH 3, the protein solubility was around 20% for both controls and alkaline pH shifts with and without salt, whereas the protein content of 2 g/L reached higher values.When the pH was adjusted Fig. 2. Ratio of D-isomers in 10 amino acids for supernatants of 10 g/L protein suspensions raised to different pH levels and back to pH 7, n = 3. Extract refers to the alfalfa protein extract without being dissolved prior to analysis.Serine for the extract was under the detection limit.Fig. 3. Cross-linked amino acids represented by groups of lanthionine (LAN), carboxymethyllysine (CML), carboxyethyllysine (CEL), lysinoalanine (LAL) and furosine (ng/mg protein) for supernatants of 10 g/L protein suspensions raised to different pH levels and adjusted back to pH 7, n = 2. Extract refers to the alfalfa protein extract without being dissolved prior to analysis.LAL for the extract were under the detection limit.Bars with different letters within the same group are significantly different (P < 0.05).
to pH 2 followed by adjustment to pH 3 (pH 2 → 3) for the 2 g/L protein concentration, a higher solubility was observed for the samples without addition of salt reaching 40% for the control and 60% for the alkaline pH shift.For the 10 g/L protein concentration, a solubility of around 20% was observed for all the samples with pH 2 → 3 treatment.In general, the conductivity of the 10 g/L protein solutions with no added salt (Fig. 5D) reached higher values compared to the 2 g/L solutions (Fig. 5C), indicating that the alfalfa protein extract itself contributes to the ionic strength due to impurities as salts.This might explain the lower solubility observed for the 10 g/L solutions, especially for the alkaline pH shift at pH 2 → 3 and 7. Thus, a dialysis would be interesting to investigate in order to achieve higher solubilities.Betschart (1974) also observed a lower solubility with higher protein concentration for acid precipitated alfalfa protein, which likewise was due to a higher ionic strength.
Regarding the conductivity, it can be observed that the values are significantly different for most of the pH-values.Optimally, similar conductivities for the samples with and without addition of salt would ease the comparison of solubilities.However, the differences are caused by different amount of HCl needed to adjust to pH 3, 5 and 7, respectively, after the alkaline pH shift and likewise for the pH 2 → 3 treatment.This is especially expressed at 10 g/L as the alfalfa protein extract has a high buffer capacity and thus further adjusting is needed to the different pH-values, creating bigger differences in conductivity.
Considering the addition of salt, Wang and Kinsella (1976) discovered a negligible effect of 0.05-0.2M NaCl on the solubility, which is comparable with our results, except pH 2 → 3 at low protein concentration and pH 7, where the most significant differences were observed.Lamsal et al. (2007) obtained similar protein solubility as the presented controls with a solubility of 20-40% from pH 2 to 9 measured on acid precipitated alfalfa.Other studies using alkaline extracted and membrane filtrated alfalfa protein have discovered a higher solubility at pH 2 reaching around 40-60% solubility (Hojilla-Evangelista et al., 2017;Knuckles & Kohler, 1982).However, no one has previously documented the effect of adjusting the pH to 2 followed by pH 3, which in the case of a lower protein concentration resulted in a higher protein solubility.For    Kohler, 1982;Lamsal et al., 2007).The alkaline treated sample at pH 7 showed solubility comparable to that of ultrafiltrated alfalfa concentrate (Lamsal et al., 2007), whereas purified rubisco has been documented to reach even higher solubility of 90% (Martin et al., 2019).Thus, the lower protein solubility in this study can be due to interference from impurities in the protein powder such as salts and polyphenols.Moreover, a lower protein concentration resulted in higher solubility especially for the pH 2 → 3 treatment and for pH 7, indicating that the solubility is affected by the amount of protein molecules or other components present in the powder.
It must, however, be noted that a lower centrifugation force was applied to remove insoluble protein in the present study compared to other studies and this might result in a higher protein content in the supernatant (Hojilla-Evangelista et al., 2017;Lamsal et al., 2007).Nevertheless, it can be observed that pH shift improved the solubility significantly at pH 7 (2 and 10 g/L) and at pH 2 → 3 (2 g/L) in comparison to solubilization in buffer.Jiang et al. (2010) as well discovered a significant improved solubility on soy protein with acidic and alkaline pH shift where the alkaline pH shift caused the biggest improvement.This was mainly caused by the protein structural unfolding and cleavage of some disulfide bonds, creating further repulsive electrostatic interactions and hence a higher stability of the solution.Thus, the pH shift creates further possibilities for usage in different food systems in relation to the solubility.

Protein composition
The protein composition was evaluated by SDS-PAGE for the control and alkaline pH shift samples with 2 g/L protein without addition of salt at pH 2 → 3, 3, 5 and 7 (Fig. 6).Overall, more protein could be detected for alkaline pH shift readjusted to pH 2 → 3 and pH 7 compared with the control samples, supporting the results from the protein solubility.However, a higher extent of smear could be observed in these samples indicating that some protein might be hydrolyzed and cross-linked due to the alkaline pH shift.From the rubisco standard, a band around 50 and 15 kDa could be detected reflecting the large (50-55 kDa) and small subunit (12-18 kDa) (Andersson & Backlund, 2008).The large subunit could be detected in the controls for pH 2 → 3 and pH 3, whereas for the alkaline pH shift this was not observed probably because of cross-linking.The small subunit could be detected in most samples expect pH 5, where no distinctive bands were observed due to the pH being close to the isoelectric point of rubisco.For all of the samples, except pH 5, a band above 260 kDa could be observed indicating a content of cross-linked proteins as already discussed (Fig. 3).Other bands were also observed in the different samples between 15 and kDa as well as in the rubisco standard, indicating degradation of the protein.However, some of the bands could as well be due to other membrane and metabolism relevant proteins in the extract (Barbeau & Kinsella, 1988;Lamsal et al., 2007).

Zeta-potential
The zeta potential was evaluated for the controls and alkaline pH shift samples with 2 g/L protein with (185 mM NaCl) and without addition of salt at pH 2 → 3, 3, 5 and 7 (Fig. 7).The zeta potential describes the surface charge of the proteins in the given sample and thus, can indicate stability of the proteins in solution.The highest value was observed for pH 7 without salt reaching around − 25 mV, followed by pH 3 with around 20 mV, pH 2 → 3 with 15 mV and pH 5 with around − mV.The obtained zeta potentials reflected the isoelectric point of Rubisco at pH 4.5 (Bahr, Bourque, & Smith, 1977) as the measurements crossed zero mV between pH 3 and 5.In general, the alkaline pH shift reached slightly higher values compared with the control except for pH 7. All of the samples with addition of salt had a zeta potential close to zero mV.However, as samples at pH 7 with addition of salt showed a protein solubility reaching around 50% for the control and around 70% for the alkaline pH shift (Fig. 5), the salt level might interfere with the measurements.The zeta potential is measured by the electric potential in the outer layer of the molecules and as the Na + Cl − ions will interact with the protein, it affects the charge and hence the measurements (Garg, Cartier, Bishop, & Velegol, 2016).Nevertheless, the relationship between the different pH-levels with addition of salt were comparable to the samples without salt but in a much smaller scale.For the samples without addition of salt, the solubility of the different samples were more or less reflected on the zeta potential.However, pH 3 resulted in a higher zeta potential compared to pH 2 → 3 and 5, when in fact the latter resulted in a higher protein solubility.This might be due to that the measurements were taken after 30 min of settlement reflecting the properties of the soluble system instead of the whole system.Furthermore, pH 2 → 3 resulted in a higher ionic strength than pH 3 (Fig. 5C), Fig. 6.SDS-PAGE with alfalfa protein solutions.Lanes from left to right: Molecular weight marker with mass (kDa) indicated to the left of the lane, rubisco standard (1 g/L), supernatants obtained after centrifugation of controls and alkaline samples at the indicated pH-levels with a starting protein concentration of 2 g/L.which could reflect the lower zeta potential.Overall, samples at pH 7 without addition of salt showed the highest zeta potential and might thus be attributed as the most stable solution of the analyzed samples.

Foam overrun and stability
The foaming properties of the alfalfa protein were evaluated for the controls and alkaline pH shift samples with 2 g/L protein with (185 mM NaCl) and without addition of salt at pH 2 → 3, 3, 5 and 7 b y foam overrun (foam capacity) and foam stability (Fig. 8).Foam overrun is related to the proportion of interfacial area that the protein film can cover, whereas the foam stability defines the capability of the protein to stabilize foam over time (Damodaran, 2017).For the foam overrun, there was no interaction between salt content and the solution media, whereas the other interactions (pH and salt, pH and solution) had a significant impact (P < 0.05).The highest foam overrun (⁓140%) was observed for pH 2 → 3 with buffer and after alkaline pH shift (Fig. 8).Alkaline pH shift with and without salt resulted in the highest overruns at pH 3, 5 and 7 with values approximately twice as high compared to the controls.At pH 3, the alkaline pH shift reached the highest value (120%), whereas at pH 5 and 7 alkaline pH shift with addition of salt resulted in the highest overruns of 110 and 137%, respectively.Hojilla-Evangelista et al. ( 2017) using alkaline extracted alfalfa protein obtained similar tendencies as the controls, where the highest overrun was observed for pH 2, whereas pH 7 was 6 times lower.Foam overrun is greatly affected by protein concentration, and the majority of proteins reaches maximum foam overrun at protein concentration from 20 to 80 g/L (Damodaran, 2017), where this study only used 2 g/L protein.Thus, higher protein concentrations should be investigated to see how this affects the foam volume.
The effect of salt in foam produced with alfalfa extracts has not yet been documented.In this study, salt addition resulted in higher overruns for pH 5 and 7 for both control and alkaline pH shift compared to samples without salt.This might be due to neutralization of charges by the salt ions favoring protein interactions (Damodaran, 2017).Furthermore, a higher solubility was found at pH 5 with the addition of salt (Fig. 5A), which could explain the higher foam formation.For pH 7, on the other hand, a lower solubility was observed with addition of salt and thus other factors might enhance the foam formation as protein interactions caused by divalent cations (Damodaran, 2017).For pH 2 → 3 and 3, the opposite occurred, decreasing the foam overrun with addition of salt for both control and alkaline pH shift, which might be explained by the lower solubility observed for pH 2 → 3 at 2 g/L with addition of salt.
As foaming properties are related to the surface activity, correlations might be drawn to the zeta potential.For the zeta potential, the controls and alkaline pH shift samples without salt were quite similar at the different pH-values (Fig. 7), whereas the foam overrun resulted in relative different values between these samples.Thus, this indicate that there is big difference between the controls and alkaline pH shift samples that may be attributed to other factors than those of protein surface charge, which are given by the zeta potential.This could for instance be the solubility as already stated, surface hydrophobicity and air-water adsorption (Damodaran, 2017).Furthermore, the zeta potential is measured on the soluble protein, whereas the foamability also can benefit from unsolubilized protein by forcing it into the foam interface, where un-soluble protein may have a higher affinity for the air phase by a more hydrophobic nature, and thus, these measurements are not correlated directly.
Foam stability was evaluated over a 1 h period and results are shown in Fig. 9. Considering the stability of foams, the controls without salt declined rapidly, reaching nearly 0% after 15 min (Fig. 9A).The pH 2 → 3 treatment (control) without salt showed a higher stability reaching 20% after 30 min and 10% after 1 h.For the controls with addition of salt (Fig. 9B), a slightly higher stability was observed reaching 1.5-20% after 1 h.For the alkaline pH shift without salt (Fig. 9C), pH 7 showed the highest stability with >60% foam after 1 h, whereas pH 5 had 23% and pH 2 → 3 and pH 3 had 13-15% foam standing.Lastly, the alkaline pH shift with salt (Fig. 9D) resulted in the highest stability for pH reaching ⁓35%, whereas the other pH-values had ⁓15% foam remaining after 1 h.Common for other studies is that the highest stability is observed near the isoelectric point at pH 4.5 due to only slight electrostatic repulsion resulting in higher protein-protein interactions (Knuckles & Kohler, 1982;Lamsal et al., 2007;Martin et al., 2019).The poor foam stability observed in this study for the majority of the different treatments especially near the isoelectric point, could be due to presence of impurities in the protein concentrate, such as polyphenols, fibers, chlorophyll and fats that affects the stability.Hojilla-Evangelista et al. ( 2017) obtained high stability of foam produced at pH 2 after min, where the pH 2 → 3 in this study as well was higher after 15 min (70%) but decreased drastically after.Foam created at pH 7 with alfalfa protein concentrate in other studies (Knuckles & Kohler, 1982;Lamsal et al., 2007) has shown to exhibit poor foam stability, except for rubisco isolates that reaches higher stability (Martin et al., 2019).As the alkaline pH shift at pH 7, without the addition of salt, in this study resulted in a significantly higher stability, this might be relevant for creating foams with alfalfa concentrates used in neutral foods.Likewise, Liang and Kristinsson (2007) discovered higher foam overrun and stability with acid and alkaline pH shifts of egg albumen, which were due to a higher surface hydrophobicity of the proteins as a result of unfolding.Higher surface hydrophobicity results in a more rapid adsorption to the interface and re-orientation of the hydrophobic and hydrophilic side chains, which increases the ability of proteins to interact to build the protein film and hence increases the foamability (Damodaran, 2017).Thus, using either acid or alkaline pH shift increases the foamability significantly, whereas the stability needs additional investigation to create foams at different pH-levels.

Gelation
The gelation properties of the protein was evaluated for different treatments (pH shift/control, pH, salt, protein concentration).Gelation describes the ability of proteins to transfer from a liquid state to a semisolid state, which is based on different interactions between the proteins creating a continuous network that binds water molecules (Damodaran, 2017).Overall, three different tendencies in the gelation curves were observed with variations in G' values (Fig. 10).For pH 7, all samples started at a low G' value and increased drastically during heating, whereas samples at pH 2 → 3, pH 3 and pH 5 started at a higher G' value, all displaying an initial viscous character, and less dramatic increase in G' upon heating.Based on the presented data on solubility, samples at pH 2 → 3, pH 3 and pH 5 all had lower solubility and lower zeta potential than pH 7. Suspended undissolved particles and increased aggregation at pH values close to pI, may explain this observed difference in the initial viscosity compared to pH 7.However, similar for all samples were a larger increase in G' during the cooling step compared with the heating step.This indicates, that the network is highly based on hydrophobic and electrostatic interactions and hydrogen bonds (Martin et al., 2014).
Gelation of 36 g/L alfalfa protein samples resulted in a low storage modulus (G') for controls with and without salt at pH 2 → 3, 3 and 7, whereas pH 5 reached a G' of 110-115 Pa (Table 3).For the alkaline pH shift samples, the strongest gel at 36 g/L protein concentration was observed at pH 7 reaching 961 Pa.For pH 5 and pH 7, exhibiting the highest gel strength, the protein concentration was raised to 72 g/L.This resulted in stronger gels for both controls and samples subjected to alkaline pH shift.The strongest gel formation was the alkaline pH shift adjusted to pH 7 reaching 2584 Pa, which was a 6 times stronger gel compared to the control.Thus, the alkaline pH shift shows a great effect on gelation with high protein concentration.
As a high concentration of NaOH was needed to obtain the target pH due to the higher protein concentration, the lowest ionic strength possible for the alkaline pH shift was 200 mM for the samples with a protein concentration of 36 g/L and 400 mM for the samples with 72 g/ L. The controls on the other hand were solubilized in buffer systems with an ionic strength of 15 mM and with the addition of NaCl (185 mM), adding an ionic strength of 200 mM for both 36 and 72 g/L.These were subsequently adjusted to reach the desired pH.Thus, the actual ionic strength is higher as previously observed with the solubility (3.2), as the alfalfa protein extract itself increases the ionic strength due to content of salts.
For the addition of salt, results indicate a slightly weaker gel formation in the controls, but this was not significant, which might be due to a smaller difference in ionic strength due to the higher amount of protein powder already containing salts.Weaker gelation upon addition  of salt has previously been documented for alfalfa protein contrasting other proteins as pea protein and whey protein isolate, where the G' increases (Knuckles & Kohler, 1982;Lamsal et al., 2005;Martin et al., 2014).Thus, the addition of salt to the hydrophobic alfalfa protein might increase the interactions, resulting in fast aggregation and hence reduce the network formation.
Gelation is a network formation and is thus, greatly affected by the balance between attractive and repulsive forces (Damodaran, 2017).As the zeta potential describes the degree of repulsive interactions, these measurements may relate to each other.However, the solutions used for gelation had a higher protein concentration and therefore a higher ionic strength would be expected (Fig. 5), giving a lower zeta potential with the same trends.For pH 2 → 3 and 3, both controls and samples subjected to alkaline pH shift, exhibited poor gelling properties and likewise had similar zeta potentials.Weak gels are often formed at extreme pH due to strong repulsive interactions, even though the zeta potential of these samples did not result in high zeta potentials (±30 mV) (Damodaran, 2017).Instead, the poor solubility of these samples might be the cause of the weak gel formation.For pH 5 and 7, the control and alkaline pH shift samples resulted in different gelling properties, where the control at pH 5 resulted in the strongest gel and the alkaline pH shift sample resulted in the strongest gel at pH 7. For the zeta potential, the control and alkaline pH shift samples reached similar negative values for pH 5 and for pH 7. Thus, the differences in these samples must be attributed to other factors than what is reflected by the zeta potential.Solutions at pH near pI often result in particulate gel formation, whereas solution at pH > pI > pH result in fine-stranded gels (Foegeding, 2006), which may be part of the observed difference between the gelation at pH 5 ⁓ pI and pH 7 > pI.Furthermore, the higher solubility at pH 7 may improve the gel formation, which is greatly affected by the alkaline pH shift.Liang and Kristinsson (2007) discovered a higher surface hydrophobicity of egg albumen subjected to acidic and alkaline pH shift refolded to pH 8.5 due to the structural unfolding of the protein.In connection, a higher surface hydrophobicity has been found to correlate to improved gel texture for egg albumen due to stronger hydrophobic interactions (Hammershøj, Rasmussen, Carstens, & Pedersen, 2006).Hence, this could explain the remarkably stronger gelation for the alkaline pH shift at pH 7.
In this study the stronger gel formation for the controls was observed at pH 5, whereas Knuckles and Kohler (1982) discovered weaker gels below pH 7.This creates possibilities for gelation of acidic foods with alfalfa protein.The alkaline pH shift resulted in stronger gels for pH 7 compared to the control, creating possibilities for gelation at lower protein concentration.In general, the critical gelling concentration of alfalfa protein is found much lower compared to whey, egg white and soy protein where more protein is needed for gelation (Martin et al, 2014(Martin et al, , 2019)).Thus, there is a great potential for the use of alfalfa protein subjected to alkaline pH shift in food products that only need a mild processing in relation to gelation temperature and hence being a more sustainable option.

Conclusion
This study evaluated the functional properties of alfalfa protein when subjected to a pH shift process or by suspending in buffer solutions in terms of solubility, foamability and gelation.The protein quality of amino acids in relation to the alkaline pH shift was investigated with different alkaline pH-values.A pH of 11 was chosen as this was found to increase the solubility without inducing major cross-linking and amino acid racemization as observed for pH 12.It was found that the alkaline pH shift (pH 11) was very effective as this improved the solubility, foamability and gelation, especially at pH 7. Subjecting the proteins to an acidic pH shift (adjusting from pH 2 → 3) improved the foaming overrun and the solubility at low protein concentration (2 g/L).Thus, it can be concluded that using either acid or alkaline pH shift processes improves essential functionalities in comparison to solubilization in buffer.Addition of salt (NaCl) was also investigated, which mainly decreased the solubility for the alkaline pH shift at pH 7 (2 g/L and 10 g/ L) and pH 2 → 3 (2 g/L), whereas an increase was observed at pH 5 for both control and alkaline pH shift at 2 g/L.Addition of salt, furthermore improved the foaming overrun for pH 5 and 7 for both controls and alkaline pH shift.These findings create further possibilities for the usage of alfalfa protein as a food ingredient, where it can provide functional properties of foaming and gelation.However, future investigations in terms of taste and digestibility would be necessary in order to confirm the potential of alfalfa protein as an alternative protein source.
program for the bioeconomy of the region of Midtjylland" for financial support of the present work.

Fig. 1 .
Fig. 1.Solubility (%) of 10 g/L alfalfa protein suspensions exposed to different pH-values of the alkaline pH shift and readjusted to pH 7, n = 6.Data-points with different letters are significantly different (P < 0.05).

Fig. 4 .
Fig. 4. Effect of NaCl concentrations and time of NaCl addition on protein solubility.Addition of salt (0, 50, 100 and 150 mM) before and after alkaline pH shift (raised to pH 11 and back to pH 7) with an alfalfa protein concentration of 10 g/L, n = 6.Bars with different letters are significantly different (P < 0.05).

Fig. 5 .
Fig. 5. Protein solubility (A-B) and conductivity (C-D) for 2 g/L (A and C) and 10 g/L (B and D) alfalfa protein suspensions at different pH-values in buffer solution (controls) and with alkaline pH shift before final pH adjustment with (185 and 150 mM salt for 2 and 10 g/L, respectively) and without salt, n = 6.Bars with different letters in the same graph are significantly different (P < 0.05).

Fig. 7 .
Fig. 7. Zeta potential of 2 g/L alfalfa protein suspensions at different pH-levels in buffer solutions (controls) and with alkaline pH shift with (185 mM) and without salt, n = 4. Bars with different letters are significantly different (P < 0.05).

Fig. 8 .
Fig. 8. Relative foam overrun of alfalfa protein as function of pH-value, solution media and salt level.Volume of foam created compared to the starting solution of the different samples at 2 g/L protein content, n = 2. Bars with different letters are significantly different (P < 0.05).

Fig. 9 .
Fig. 9. Relative foam stability given as foam volume (%) as function of time for the different samples at 2 g/L protein content.A) Control, B) Control with 185 mM salt, C) Alkaline pH shift and D) Alkaline pH shift with 185 mM salt, n = 2.

Fig. 10 .
Fig. 10.Examples of thermal gelation profiles of alfalfa protein with storage modulus (G') and temperature as function of time showing three different tendencies of curves dependent on pH.Example pH 7 (shown: alkaline pH shift, 36 g/L protein), example pH 5 (shown: control, 36 g/L protein, 15 mM) and example pH 2 → 3 and 3 (shown: control, pH 3, 36 g/L protein, 200 mM).

Table 1
The optimized precursor ions, and MRM transitions given a quantifier and qualifier ions for analytes.

Table 2
Protein solutions as function of solution media, pH, protein concentration and salt level resulting in different ionic strengths.