Synergistic effects of processing parameters on the biochemical and physical properties of tofu made from yellow field pea (Pisum sativum), as determined by response surface methodology

Abstract Peas are an underutilized crop that do not require allergen labeling and are rarely genetically modified. Peas contain less protein than soy and vary in protein composition. Because peas contain more starch than soy and less lipids, an alternative procedure for pea tofu production needs to be developed to prevent excessive starch gelatinization while promoting curd development. To accomplish this, a response surface model design was utilized to determine optimal oil addition, cook time, and salt concentration. Treatment ranges were from 0.0% to 4.2% for oil addition, 60–134 min for cook time, and 5.0%–9.2% for MgCl2 addition. Treatments had varying effects on tofu texture. Cook time was directly proportional to the hardness and could be used to match the soft, firm, and extra firm texture targets of conventional soy tofu. Protein secondary structure was not related to gel strength, indicating a system with synergies between multiple components other than protein. This research will help satisfy the growing demand for alternatives to soy‐based foods.

Because proteins from the yellow field pea (Pisum sativum) have similar biochemical composition to that of soy, it can be hypothesized that they will be able to coagulate, make a curd, and form a soy-free tofu analog.
While tofu production can vary by manufacturer, the basic steps are soaking and grinding soybeans into a slurry, filtering the slurry to make soy milk, heating the milk, adding coagulants, and pressing the resulting curds. In traditional tofu, soy proteins need to dissociate to expose the hydrophobic regions. This occurs between 65 and 75°C depending on ionic strength, with higher ionic strength increasing the temperature required for dissociation of glycinin and β-conglycinin in soy (Matsudomi et al., 1985). Other research found that some of these proteins can remain undissociated after 30 min (Wu et al., 2017), and pea proteins are more thermally stable, dissociating between 75 and 85°C (Mession et al., 2015). Because the change in dissociation temperature is dependent on ionic strength (Matsudomi et al., 1985), the degree of dissociation prior to coagulant addition will affect the amount of protein available to react with the coagulant. Unless the proteins are fully dissociated at the time of coagulant addition, excess coagulant is likely to result (Cai & Chang, 1998).
While there have been many studies on soy tofu and the underlying mechanisms required for its production, there has been very little published on production and quality of soy-free tofu. The most comprehensive study reviewed five beans, chickpeas, lentils, smooth peas, mung beans, and fava beans (Cai et al., 2002). The methods called for extensive extraction and centrifugation, followed by an aggressive heating step, requiring 10 min of boiling. A study by Jayasena et al. (2010) looked to replace a certain amount of soy beans with lupin beans for lower fat and cost tofu. Up to 40% of the soy beans could be replaced with no significant effects to texture. Gebre-Egziabher and Sumner (1983) produced tofu with pea flour and pea protein concentrate using calcium sulfate and found that an acceptable curd could be formed. However, the aforementioned studies did not look at industry-friendly methods of tofu production.
In work by DePalma et al. (2019), a physical disruption of curds made from the coagulation of yellow field pea proteins with heat and MgCl 2 produced a tofu analogue with similar characteristics as commercially available soy tofu. This study investigated quality, protein secondary structure, and microstructure of pea-based tofu analogues in response to lipid level and physical disruption of protein curds followed by repressing. It also focused on individual variables and not the synergistic effects of multiple variables on tofu biochemical properties and resulting textures. Considering the affects ionic strength, the temperature required for dissociation, time-temperature effects on protein dissociated, and the effect of oil concentration on protein dissociation and aggregation, there is an ample amount of synergistic interaction in the system. The objective of this study is to determine the synergistic effects of different processing factors (cook time, MgCl 2 , and oil concentration).

| Materials
Yellow split peas were obtained from Columbia Bean & Produce. The dried, refined Japanese Nigari (MgCl 2 ) was purchased from Handy Pantry.

| Response Surface Methodology
A variation of response surface methodology (RSM) used by Smith et al. (2019) and Smith et al. (2012)  where Y is the response variable. When i = j, β 0 is the coefficient of intercept, β i is the coefficient for linear, β ij is the coefficient for quadratic, and when i ≠ j, β ij is the cross-product coefficient. x i and x j are terms used to describe the independent factors.
The sequential model sum of squares (SMSS), lack-of-fit tests, and the multiple correlation coefficient (R 2 ) were used to select a model (mean = no model, linear, quadratic, or cross-product) for each response. Significance was defined as p < .05.

| Pea Milk and Tofu Preparation
Pea milk and pea-based tofu were produced as described by DePalma et al. (2019) with changes. Pea milk (481 g) plus corn oil treatments (0.0%-4.2% [w/w]) were added to a 1,000-ml beaker.
The mixture was blended with a Cuisinart immersion blender for 2 min to homogenize. The beaker was placed in a 98°C water bath for the duration of a given cook time (45-150 min). After heating, MgCl 2 (4.2%-10% [w/w]) was added over the course of 10 s, stirring constantly. After 10 min, the curds were poured into a 12.8 × 9.1 × 7.9 cm tofu mold lined with a cotton cloth. The cloth was folded once over the curds, and a perforated piece of plastic was placed on the cloth before the other sides of the cloth were folded over to prevent molding defects. A 5 kg weight was placed on top of the mold for 10 min. The pea tofu was removed from the mold and cooled for 10 min before being wrapped in plastic wrap and refrigerated.
An additional sample set of pea-based tofu was prepared in duplicate using the method previously described. The tofu was cut into strips and placed into a 50-mL centrifuge tube. The tube was submerged in liquid nitrogen until tofu was frozen. The samples were held in a −80°C freezer until being freeze-dried in a FreeZone 6 (Labconco) at −56°C and 0.02 mbar. The freeze-dried samples were stored in a −80°C freezer until biochemical analysis was performed.

| Physical Properties of Tofu
Samples were equilibrated to room temperature for 30 min prior to testing. Two blades fastened together 10 mm apart were used to cut the samples into 10 mm cubes. Tofu color was assessed by L*, a*, and b* color parameters according to the CIELAB international system of color measurement. To accomplish this, a piece of tofu larger than the 8-mm-diameter colorimeter aperture was placed on Konica Minolta Spectrophotometer CM-5 (Konica Minolta Sensing Americas, Inc.) equipped with an 8 mm aperture and instrument set to D65 with a 10% light angle. Excess moisture was blotted from the surface of the sample prior to analysis.

| Percent Yield
The mass of the block of tofu was measured on a Mettler Toledo New Classic SG digital balance. Percent yield was determined by Equation 2:

| Percent Retained Solids
Retained solids are the percent of solids that went into the system and were not pressed out. The value was calculated by the following equation:

| Optimization
Optimization of desirability was determined by the equation as described by Smith et al. (2019): where D represents desirability; d i is the desirability for each response; and r i is relative importance (with 1 being most desirable, D values can range from 0 to 1).

| FT-IR Spectroscopic Analysis
Freeze-dried samples were equilibrated to room temperature for 45 min prior to obtaining spectra on a Nicolet iS10 Fourier transform IR spectrometer (Thermo Fisher Scientific) that was purged continuously with dry air and equipped with a diamond attenuated total reflection accessory (Smart orbit, single bounce, 45°), which was used to measure the IR spectra. FT-IR spectral analysis was performed as described by DePalma et al. (2019). The spectra for added oils were subtracted prior to baseline correction and deconvolution. (1)

| Molecular Weight Analysis
Freeze-dried ground sample (0.5 g) was placed in centrifuge tubes (50 ml) and vortexed at 450 rpm for 30 min with an excess of hexane and centrifuged for 1 min at 340 g. The supernatant (lipids) was decanted, and remaining hexane was evaporated in open air. Removal of lipids improved the resolution of the electropherograms.
Extractions were carried out as described by DePalma et al. (2019) and analyzed with an Agilent Bioanalyzer 2100 (Agilent).

| Microscopy
A piece of freeze-dried sample, approximately 2 mm × 2 mm, was

| Physical Properties of Tofu
The properties (hardness, springiness, cohesiveness, and color coordinates) of the tofu were optimized using an RSM for cook time and MgCl 2 content. The linear model for hardness ( Figure 1a) was significant (p = .0165). The hardness of the tofu was significantly affected by cook time (p = .0049), increasing the hardness by 50 g between 60 and 134 min when all other variables were held constant. The hardness was not significantly affected by = MgCl 2 content (p = .3175), reducing the hardness by 10 g between 5.05% and 9.15% (w/w) when all other variables were held constant. The hardness was not significantly affected by oil addition (p = .1397), decreasing the hardness by approximately 20 g between 0.62% and 3.55% (w/w) oil.
The quadratic model for springiness ( The cook time is the only significant factor in determining the hardness, and one of two significant factors affecting cohesiveness. With heating, soy proteins have been shown to not immediately dissociate, with some remaining undissociated after 30 min of heating (Wu et al., 2017). This phenomenon could be exacerbated in pea proteins, which require 75-85°C to dissociate, as opposed to soy F I G U R E 1 A graphical representation of the effect of present MgCl 2 (w/w) (y-axis) and time (x-axis) at 2.14% (w/w) oil addition on the physical properties of pea-based tofu with a) hardness, b) cohesiveness, and c) springiness on the vertical axis (top row) and tofu color, d) L*, e) a*, and f) b* on the vertical axis (bottom row) that only requires 65-75°C to dissociate (Mession et al., 2015;Wu et al., 2017).
The MgCl 2 content was not a significant factor for any mechanical property tested. However, the rate of coagulant addition has been shown to affect the mechanical properties of soy tofu (Cai & Chang, 1998). In this study, the concentration of MgCl 2 was adjusted by changing the amount of the stock MgCl 2 solution. While the duration of MgCl 2 addition was constant for all samples (10 s), the amount was not. This means the rate of MgCl 2 addition varied between treatments. This could potentially have led to confounding variables but was unavoidable given the needed treatment variables.
The addition of oil did not significantly affect the hardness, which varied from previous work that showed a significant decrease In these data, the positive a* indicates a very slight redness, but the predominating color is yellow, as indicated by the b* values.
Tofu prepared from yellow split peas had a more yellow appearance than traditional soy tofu, which could be considered a product defect (Kim and Wicker, 2005). Fortunately, the yellow hue can be mitigated through the addition of oil and modification of heat treatments (Figure 1a-f). For reference, Kim and Wicker (2005) reported L* values of 90 and 93 for soy tofu, while the pea tofu in this study ranged from 82 to 85.

| Percent Yield
The (w/w) addition levels. The percent yield was most affected by oil addition, which is to be expected, since the oil is incorporated into the gel and the bulk of the water was removed during pressing.

| Percent Retained Solids
While percent yield is important, greater attention must be given to the percent retained solids. This value represents additional nutrients originating from the raw peas that get incorporated into the tofu gel, resulting in less loss for the manufacturer. (w/w) between 0.62% and 3.55% (w/w).
It is important to note that retained solids accounted for only the solids from the pea milk and MgCl 2 . Since the oil is completely incorporated by the curds, the only solids that can be removed during pressing come from the pea milk and MgCl 2 . Therefore, it is important to focus on how much of the pea solids are lost and MgCl 2 is left, because it is difficult to determine how much of this salt is incorporated. The oil addition had little effect on retained solids at the lower levels. However, after a retained solids value of 2.4% (w/w) was reached, the values increase rapidly, as it was a quadratic function.
This likely represents a critical value of when proteins with higher binding affinities to lipids become saturated and excess oil is free to bind to unassociated protein subunits.

| Optimization
For this research, the optimal quality tofu was one considered to

| FT-IR Spectroscopic Analysis
FT-IR spectroscopy can provide information on protein structure in pea tofu. The range of 1,700--600 cm −1 , or amide I band, was decon- alone. However, the interaction between MgCl 2 and oil addition (p = .0162) did have a significant affect; at low levels of oil addition, the percent of β-sheets increased from 7% to 14% between the MgCl 2 concentration of 5.0 and 9.2% (w/w) and the maximum cook time. At high levels of oil addition, β-sheets decreased from 13% to 10% between the MgCl 2 concentration of 5.0 and 9.2% (w/w) at the maximum cook time. This may be related to ionic-mediated dissociation of vicilin and legumin proteins (Matsudomi et al., 1985).
For this reason, the dissociation of the proteins was likely related to the low percent of β-sheets in solutions of low ionic strengths.
Since β-sheets correlate with retained solids, it is possible that the solids being retained were high in β-sheets and would thus be protein.

F I G U R E 4
A graphical representation of the effect of percent MgCl 2 (w/w) (y-axis) and time (x-axis) at (a) 0.62% (w/w), (b) 2.14% (w/w), and (c) 3.55% (w/w) oil addition with the vertical axis being α-helices (top row) and the effect of percent MgCl 2 (w/w) (y-axis) and time (x-axis) at (d) 0.62% (w/w), (e) 2.14% (w/w), and (f) 3.55% (w/w) oil addition with the vertical axis being β-sheets (bottom row) F I G U R E 5 A graphical representation of the effect of percent MgCl 2 (w/w) (y-axis) and time (x-axis) on surface hydrophobicity (µg of bound bromophenol blue) (vertical axis) with oil at a) 0.620%, (b) 2.14%, and (c) 3.55%

| Surface Hydrophobicity
The quadratic model for surface hydrophobicity (Figure 5a-c) was significant (p = .0074). The surface hydrophobicity of the tofu was not significantly affected by cook time (p = .3269). The surface hydrophobicity was significantly affected by the MgCl 2 content (p = .0498). The cook time and MgCl 2 content did have an interaction with oil levels; at low oil concentrations, surface hydrophobicity was directly proportionate to cook time and MgCl 2 concentrations, but at high oil levels the surface hydrophobicity was inversely proportionate to cook time and MgCl 2 concentrations ( Figure 5c). The surface hydrophobicity was significantly affected by oil addition (p = .0038), decreasing by approximately 25-50 µg BPB between 0.62% and 3.55% (w/w) oil if all other factors were held constant.
At low ionic strength, cook time increased the surface hydrophobicity, and at high ionic strength, cook time decreased surface hydrophobicity. Lower ionic strength is known to promote aggregation of legumin (Gueguen et al., 1988), so there should be less dissociation and thus less exposure of the hydrophobic interior regions of proteins, resulting in less binding of BPB. At low oil concentrations, the cook time increased the surface hydrophobicity, but at higher oil concentrations, cook time decreased the surface hydrophobicity. As the vicilin and legumin are heated, they dissociate into subunits, decreasing molecular weight and exposing hydrophobic regions (Mession et al., 2015). With oil addition, the newly exposed hydrophobic regions should associate with the lipids. This was seen in the decrease in surface hydrophobicity as the samples were cooked longer, with higher ionic strengths and oil addition. The change in the trend from increasing to decreasing surface hydrophobicity when oil levels are increased (Figure 5a-c) is likely from newly exposed hydrophobic surfaces on the dissociated protein subunits coating the surface of oil globules (Mession et al., 2015).
In previous studies, secondary structure, α-helices and β-sheets, is often proportional to surface hydrophobicity and textural properties of tofu (DePalma et al., 2019). Within these studies, it is typical to have one or two variables and measure the treatment's effect on the resulting gel. It is often found that protein secondary structures are proportional to gel textural properties. In this study, exposure to heat, ion concentration, and oil addition were studied. It was found that treatment effects were not necessarily proportional to resulting protein secondary structure data. The findings of this study suggest that the synergy between treatments has a large influence on protein secondary structure, surface hydrophobicity, and gel strength.
This phenomenon is further compounded by the fact that in foods and other biological materials, the material being studied is rarely comprised of a single component. In this study, the treatments likely interacted with other endogenous components such as starch, polar lipids, sugars, and nonstarch polysaccharides to affect gel strength.
However, from this work, it is evident that quantification of α helices and β sheets alone may not be a good indicator of gel strength and does not necessarily have to correlate with gel textural properties.
To this end, care must be taken in subsequent studies to ensure that protein structure is not falsely correlated with gel textural properties and casual relationships must be demonstrated.

| Molecular Weight Analysis
While field peas differ from soybeans in many aspects, the major protein classes of field peas are structurally analogous to glycinin and β-conglycinin of soy. Glycinin and β-conglycinin proteins account for 70% of the total protein in soybeans and are critical to the formation of tofu gels (Poysa et al., 2006). Glycinin forms a firm gel when unfolded by heat to expose cysteine, which forms intramolecular disulfide bridges (Phillips & Williams, 2011). While glycinin has between ten and 15 cysteine residues, β-conglycinin has between two and three and forms gels with lower hardness values and more elasticity (Phillips & Williams, 2011). The legumin fraction in peas is similar to the glycinin fraction in soy, each being a 360 kDa hexamer (Phillips & Williams, 2011;Schwenke et al., 2001) comprised of acidic and basic subunits (Bacon et al., 1987). The key difference is that glycinin is held together with disulfide bonds that form interprotein cross-linkages, while the legumin fraction is held together by strong hydrophobic interactions (Phillips & Williams, 2011). In systems with high hydrophobicity, protein-protein interactions are more abundant than protein-solvent interactions resulting in coagulum rather than formation of a true gel (Phillips & Williams, 2011). However, legumin has around 63% of the cysteine as glycinin (O'Kane et al., 2004a), so disulfide bonds may still form during aggregation to form curds similar to that of soy. Furthermore, when glycinin gels have been exposed to 2-mercaptoethanol, they were only partially dissolved, indicating the importance of hydrophobic interactions (Nakamura et al., 1984). It seems likely that both glycinin and legumin gels are formed through disulfide bonds of cysteine residues and hydrophobic interactions between acidic and basic subunits. Soy β-conglycinin and pea vicilin are trimetric proteins of ~170 kDa (Maruyama et al., 1999;O'Kane et al., 2004b). Therefore, understanding the relationship between processing perimeters, pea tofu gel physical properties, and protein molecular weight distribution is critical.
The model for the ~21 kDa peak (Figure 6a) was not significant (p = .6912). The ~24 kDa peak of the tofu was not significantly af-  2b and 6b).
The linear model for the ~54 kDa peak (Figure 6c) was not significant (p = .0761), nor was it significantly affected by MgCl 2 (p = .4190) or oil addition (p = .5192). However, the ~54 kDa peak area was significantly affected by time (p = .0170), decreasing by approximately 1.5% (w/w) between 60 and 134 min when all other variables are held constant.
One of the key proteins in peas is legumin. The ~21 and ~24 kDa polypeptides are two of the alkaline polypeptides that make up a legumin subunit. Legumin breaks down into subunits that are reported to be approximately between 54 and 60 kDa (Schwenke et al., 1990).
The breakdown of the ~54 kDa protein into the ~21 and ~24 kDa is consistent with glycinin of soy breaking into subunits with heating during soy tofu production. These data also indicate that pea-based tofu follows a similar mechanism of formation as conventional soy tofu as outlined above. This hypothesis was further supported by the surface hydrophobicity data that demonstrated an increase in dissociation with increased cook time (Figures 5a-c and 6c). The amount that cook time increased the ~21 kDa and the ~24 kDa peak areas increased with oil addition. However, oil addition did not affect the ~54 kDa peak as much, indicating that legumin's breakdown was not augmented by oil addition, but the retention of ~21 and ~24 kDa polypeptides in the pea tofu was.
The other critical protein in the pea tofu system was vicilin, which is comprised of subunits of ~33, ~50, and ~71 kDa. The quadratic model for the ~33 kDa (Figure 6d) was not significant (p = .3262).
The ~33 kDa of the tofu was not significantly affected by cook time (p = .4447), MgCl 2 (p = .6695), or oil addition (p = .4023). The change in the ~33 kDa was difficult to interpret, as the ~50 kDa subunit is comprised in part of a polypeptide that is also ~33 kDa. The quadratic model for the ~50 kDa ( Figure 6e) was significant (p = .0334).
The ~50 kDa peak of the extracted pea tofu proteins was not significantly affected by cook time (p = .8635), MgCl 2 (p = .4259), or oil addition (p = .2310) alone. However, the combined effect of the MgCl 2 and oil addition was significant (p = .0178), as well as the quadratic effect of the oil addition (p = .0089). The ~33 and ~50 kDa peaks follow the same trend as the ~21 and ~24 kDa peaks, where cook time increased peak area more at high oil addition. The linear model for the ~71 kDa ( Figure 6f) was not significant (p = .0603). The ~71 kDa of the tofu was significantly affected by cook time (p = .0244), reducing the ~71 kDa by 2% (w/w) between 60 and 134 min when all other variables were held constant. The ~71 kDa was not significantly affected by the MgCl 2 (p = .1228), reducing the ~71 kDa by 2% (w/w) between 5.05% and 9.15% (w/w) MgCl 2 addition level when all other variables were held constant. The ~71 kDa protein was not significantly affected by oil addition (p = .5415). The decrease in the ~71 kDa, a subunit of vicilin, was inversely proportionate to time and thus hardness, likely indicating that the ~71 kDa subunit was associating with other proteins during the formation of pea tofu gel.

| Microscopy
All but two SEM images (Figure 7a,d) displayed honeycomb structures, a common structure reported in soy tofu (DePalma

| CON CLUS ION
The treatment plan demonstrated a range of textures for peabased tofu could be obtained by varying cook time, MgCl 2 , and oil content. Cook time was the most important factor for determining hardness, the most important quality characteristic of tofu.
Furthermore, it was determined that the synergistic effects between treatments influenced gel strength and protein secondary structure. However, it was found that gel strength did not necessarily have to correlate with any one factor. The biochemical assays showed that there is thermal dissociation of legumin and vicilin, which forms a matrix that is affected by hydrophobic interactions. These interactions allowed for the incorporation of lipid to make a visibly smoother, lighter colored pea-based tofu analog.
These data can act as a template by manufacturers looking for a low-cost, high-protein, soy-free food.

ACK N OWLED G M ENTS
The authors would like to acknowledge the University of Idaho,

College of Agricultural and Life Sciences, and the College of Natural
Resources for financial support of the Agilent 2100 Bioanalyzer and FTIR spectrometer, respectively. The authors of this manuscript have no conflict of interests to report.