The effect of whey source on heat-induced aggregation of casein and whey protein mixtures of relevance to infant nutritional product formulation

Sweet and, to a lesser extent, acid whey protein ingredients can be used for the formulation of infant nutritional products. Unlike acid whey, sweet whey contains caseinomacropeptide (CMP), a heat-stable peptide liberated from κ-casein during cheese and ren-net casein manufacture. Four protein systems—sweet whey (SW) and acid whey (AW), with or without standardization for CMP protein content—were added to skim milk (50/50, wt/wt) and unheated or heated to 85 or 110°C. These 12 samples were assessed for physicochemical stability in the presence of added calcium at pH 6.8. The effect of CMP content on the physicochemical properties of the protein systems was also assessed. Without preheat treatment, mixtures of AW and skim milk (SM) were more heat stable than SW and SM, demonstrating the effect of whey protein type on heat stability. Preheat treatment of the SW in the presence of SM significantly improved the heat stability of the resultant protein systems on subsequent heating. All of the protein systems had significantly lower heat stability with the addition of Ca, although the reduction was significantly smaller for the heated protein systems than the unheated controls. The findings can help identify heating parameters and ingredients for optimizing processing stability and physicochemical characteristics of nutritional beverages such as infant formulations.


INTRODUCTION
Whey proteins are important functional components that typically account for 60% of the total protein found in infant formula (IF).The 2 primary sources of whey proteins used commercially are sweet whey and acid whey.The 28 member states of the European Union alone produced an average of 1,936 kt of whey powder per week in 2019 (OECD, 2019).Sweet whey is a by-product of the manufacture of rennet-coagulated cheeses and rennet casein.During cheese-making, milk is pasteurized and then subjected to hydrolysis by chymosin (EC 3.4.23.4) at the natural pH of milk, cleaving caseinomacropeptide (CMP) from the casein micelle, destabilizing its structure by reducing electrostatic and steric stabilization (Calvo et al., 1995;de Kruif and Zhulina, 1996).During this process, casein forms a curd, and the whey containing the CMP is expelled to form a serum phase.Caseinomacropeptide (64 AA) is heatstable and largely electronegative peptide, even at low pH, with the nonglycosylated and glycosylated forms having isoelectric points of 3.15 and 4.15, respectively (Thomä-Worringer et al., 2006;Kreuß et al., 2009).
There has been considerable interest in the interaction between CMP and heat labile whey proteins such as β-LG.Lieske et al. (2004) found that the sugars attached to CMP were sensitive to removal on heating and particularly on acidification.Le et al. (2016) and Villumsen et al. (2015a) found that heating whey protein at elevated temperatures caused deglycosylation and weak acid hydrolysis of CMP and helped prevent storage-induced aggregation in acidic drinks via electrostatic interactions.Gaspard et al. (2020) reported that the presence of CMP, in its glycosylation form, led to pH-and temperature-dependent changes in both denaturation and aggregation of whey proteins in whey protein isolate (WPI)-CMP mixtures at low pH (4.0), with denaturation and aggregation favored when the CMP was glycosylated; however, at higher pH (>6), removal of the glycosylated moiety increased aggregation.Croguennec et al. (2014) found that although there was greater denaturation of β-LG in the presence of CMP upon heating to 85°C, the aggregation of β-LG was hindered at neutral pH, as shown by decreased turbidity and reduced hydrodynamic size.Martinez et al. (2010) found that heating CMP:β-LG (75:25) mixtures at 90°C for 5 min led to the formation of soft gels at neutral pH at a level of β-LG below its normal concentration The effect of whey source on heat-induced aggregation of casein and whey protein mixtures of relevance to infant nutritional product formulation Bernard M. Corrigan, 1 James A. O'Mahony, 2 and Mark A. Fenelon 1,2 * for gelling, demonstrating synergistic activity between the 2 proteins.
Interpreting the effect of heating on whey protein mixtures is further confounded by the presence of intact caseins; for example, β-LG can bind to κ-CN (Singh and Creamer, 1991;Guyomarc'h et al., 2003).Several studies have found that in addition to interactions of whey proteins with caseins on the surface of casein micelles, κ-CN in milk-based systems could, upon heating at >90°C and neutral pH, disassociate into the milk serum to form soluble aggregates, potentially leaving the micelle depleted of calcium-insensitive casein (Singh and Fox, 1985;Singh and Creamer, 1991).Several authors have demonstrated a chaperone effect of κ-CN, α-CN, and β-CN on the aggregation of whey proteins (O'Kennedy and Mounsey, 2006;Kehoe and Foegeding, 2011;Gaspard et al., 2017).Joyce et al. (2017) created model infant formula systems in which the whey protein component was heated separately before formulation and subsequent heat treatment in the presence of casein.Preheat treatment of the whey component led to smaller aggregates, which helped reduce the tendency for increased viscosity and heat-induced destabilization during HTST treatment.The formation of these small whey protein aggregates had been previously shown to prevent further heatinduced changes in acidic drinks (Ryan and Foegeding, 2015).However, the degree to which these aggregates form and their subsequent effects on the heat stability of mixed whey protein-skim milk systems, such as found in the present study, remain unclear.There is also a lack of comparative studies on the use of sweet or acid whey in protein systems for use in IF manufacture, even though there is a lot of knowledge regarding the heat-stabilizing effects of CMP on whey protein-skim ingredients.
In acid casein manufacture, skim milk is typically heated and acidified to the isoelectric point of casein, pH 4.6, after which the resultant curd is washed and harvested before being processed into acid casein or neutralized again and dried into various caseinates (Pisecky, 1997).Acid whey is derived from the initial wash of the curd at pH 4.6 and thus contains more minerals than sweet whey, because of the colloidal calcium phosphate being washed out of the casein curd on acidification.As acid whey contains no CMP and more minerals than sweet whey, it is reasonable to think that it would have a significant influence on heat stability in protein systems if added at approximately 60% of total protein, as relevant to IF.The effect of increasing ionic calcium on the stability of dairy-based systems, and in particular, binding of organic phosphates in the serum phase and associated lowering of buffering capacity to offset pH-induced thermal coagulation, has been demonstrated in several studies (Sievanen et al., 2008;On-Nom et al., 2012;Crowley et al., 2014).
Previous studies using whey-based model dairy systems have shown a direct correlation between protein concentration and the denaturation and aggregation of these systems.Kehoe et al. (2011) demonstrated an increase in the rate of denaturation and aggregation of WPI and β-LG solutions on heating at neutral pH but not on aggregate size.Fitzsimons et al. (2007) found that increasing protein concentration (2-12%, wt/wt) led to increases in aggregation of whey proteins upon heating at 80°C in WPI, as measured by differential scanning calorimetry, coupled with an increase in storage modulus from rheological measurements.Calvo et al. (1995) found that the rates of denaturation of individual whey proteins heated in milk to 80°C increased as the total protein concentration increased.Based on these studies, we can expect that increasing the overall proportion of whey protein, especially in the acid wheyadjusted samples (adjusted to the same whey content as CMP-containing samples) would lead to overall decreased heat stability of protein systems compared with their unadjusted counterparts.Conversely, increasing the proportion of CMP, a heat-stable protein, should increase heat stability of protein systems.Therefore, the objective of this study was to produce and evaluate protein ingredients made from different whey protein sources that could be used in the creation of ingredients for IF with increased stability to thermal processing.
This study evaluated the effect of mixtures (50/50 wt/wt) of sweet whey or acid whey and skim ingredients on the physicochemical functionality of model infant formulations with a target whey: casein ratio of 60:40.The aim was to determine the effect of whey protein ingredient type (i.e., acid whey or sweet whey) on the thermal stability of mixed whey protein-skim based model infant formula systems.The contribution of CMP to the heat-induced changes in the whey-skim mixtures was considered in the experimental design by standardizing the sweet whey samples to the same "true" whey protein content (i.e., minus CMP) as the acid whey samples.Sweet whey can contain up to ~20% CMP in total protein, whereas acid whey is free of CMP (Thomä et al., 2006).Therefore, the authors hypothesize that the presence of CMP in mixtures of whey and skim will result in significant differences in heat-induced changes.In this study, we aimed to elucidate the effect of whey protein type-sweet whey or acid whey (total protein or CMP standardized)-on the thermal properties (denaturation, heat stability, gelation, and resistance to calcium-mediated protein aggregation) of protein systems designed for IF.The findings can be used to determine the thermal processing conditions required during whey protein ingredient

Materials
Bulk raw milk was obtained on 3 separate days from the Teagasc Animal and Grassland Research and Innovation Centre (Moorepark, Ireland) and heated to 50°C before being skimmed to approximately 0.2% (wt/wt) fat using a laboratory-scale FT15 disc bowl centrifugal separator (Armfield).From these milks, batches of AW and SW were produced.

Compositional and Mineral Analysis
Three independent lots of raw milk were collected on separate days and used to produce the skim milk powder (SMP), acid whey powder (AWP), and sweet whey powder (SWP) used in preparation of the model protein systems (Figure 1).Both the raw and skim milk samples were analyzed for protein, lactose, total solids, and fat using a Bentley Fourier transform infrared spectrophotometer (Bentley Instruments Inc.).Whey (sweet and acid) and skim concentrate produced by membrane processing and raw skim milk were freezedried and subsequently analyzed for protein content using the Kjeldahl method (IDF, 1986).All whey protein samples were analyzed for CMP content using reversed phase-HPLC as described by Gaspard and Brodkorb (2019).
Mineral profile analysis of protein systems and their ultracentrifugal sera was performed using inductively coupled plasma-mass spectrometry (ICP-MS) after microwave-assisted digestion using a Mars express digester (CEM Microwave Technology Ltd.) according to the method of Camin et al. (2012).After digestion and upon dilution, the samples were analyzed using a 7700s Agilent ICP-MS (Agilent Technologies).The carrier gas was argon, and the collision gas was helium.Samples were quantitated against standards provided by Inorganic Ventures.Run performance was checked using an internal standard, and recoveries of elements were determined by use of a certified skim milk standard (ERM BD 151; Joint Research Centre for European Union).All samples gave >90% recovery of the certified standard elements except for Ca, which gave 88%.All samples were analyzed in triplicate.

Preparation of Protein Systems for Infant Formula
Production of Mineral-Depleted AW.Six liters of skim milk was heated to 50°C before being acidified to pH 4.6 using 5 M HCl.The curds and whey were carefully separated using cheesecloth, and 0.02% sodium azide was added to the whey before adjusting to pH 6.8 using 5 M KOH and refrigerating overnight at 4°C.The whey was adjusted to 22°C the following morning before centrifugation at 6,000 × g for 20 min at 20°C and final filtration through glass wool to remove any casein fines.Three liters of each sample (at ~5.5% total solids) were ultrafiltered and then diafiltered with 2 volumes of ultrapure water to give a final total solids content of approximately 2.5% with a protein concentration factor of ~4.5 using a benchtop spiral membrane with a 10-kDa molecular weight (MW) cutoff (Millipore).The whey samples were then freeze-dried.
Production of SW.Sweet whey was manufactured from the same batch of skim milk that was used for production of the AW.Skim milk (6 L) was heated to 32°C and Chymax Plus (200 IMCU/ mL, diluted 1:10 in ultrapure water; Chr.Hansen) was added to give a final concentration of 0.2 mL of chymosin/kg of milk.The sample was stirred for 1 min before being incubated for 1 h until a curd had formed.The sample was cut, held for 30 min, and then drained through cheesecloth.Finally, 0.02% sodium azide was added, the pH was adjusted to 6.8 using 5 M KOH, and the sample refrigerated at 4°C overnight.After the whey samples were allowed to return to 22°C, the pH was again checked before centrifugation at 6,000 × g at 20°C for 20 min, followed by final filtration through glass wool to remove the fines.The SW (3 L at target of 5.5% total solids) was then ultrafiltered and diafiltered with 2 volumes of water, using the same system used for the AW, providing a final total solids content of approximately 2.5% and a concentration factor of approximately 4.5, before being freeze-dried.
Preparation of the Protein Systems.The protein systems were prepared to a total protein content of 2.8% (wt/wt), by reconstitution of either whey concentrate or sweet whey concentrate with skim milk to give mixtures containing 1.4% (wt/wt) each of whey protein and skim milk protein.The designations "SW" and "AW" were used for sweet whey and acid whey protein systems, respectively, whereas "SWA" and "AWA" refer to the protein systems after adjustment as described below.The protein fraction of the skim consisted of 20% whey protein; therefore, the target final ratio of whey protein: casein was 60:40.A second set of protein systems (SWA and AWA) based on AW-and SW-skim milk mixtures were prepared where the whey protein content was targeted to be 20% higher, at 3.1% (wt/ wt) protein, giving a final whey protein: casein ratio of 65:35.The whey components of the AWA and SWA protein systems were adjusted upward on a total protein basis to allow for the CMP content of the SWA protein system and to help standardize the AWA and SWA systems for total protein.
Heat Treatment of the Protein Systems.Heat treatment of all samples was performed at 85 or 110°C for 4 min (Figure 1) to determine the effect of heating on the whey adjusted and unadjusted systems; an unheated (i.e., control, CTRL) sample was also included for analysis.A set of 20-mL steel tubes (Elbanton B.V.) were used to hold the samples during heating in an oil bath to the required temperature under standard mixing conditions.After completing the heating step, the samples were removed from the oil bath and quickly placed on ice.The following codes, as described in Figure 1, were used to denote the samples generated based on (1) whey protein type, (2) proportion of whey protein (i.e., adjusted for CMP content), and (3) heat treatment used: CTRL (no heating), 85°C, and 110°C indicate the heat treatment temperature applied in the oil bath to each sample.The experimental design and coding used are illustrated in Figure 1.
Calcium-and Heat-Induced Protein Aggregation.Calcium stock solutions were prepared from calcium chloride dihydrate (Sigma Aldrich).These stock solutions were added (1:100 vol/vol) to the reconstituted (from freeze-dried) protein systems to give a final concentration of either 1.75 or 7.0 mM Ca, with the concentration used depending on the experiment.The concentration of Ca in the stock solutions was verified using atomic absorption spectroscopy and ICP-MS.To determine the effect of added Ca on heat stability and protein aggregation, Ca was added to one subset of the protein samples and not to the other.The samples were reconstituted as outlined in Figure 1 before adding calcium, and stirred for approximately 2 h before final adjustment of pH to 6.8.

Size and Zeta-Potential Analysis
The protein systems were reconstituted as indicated above, diluted to 1:50 (vol/vol) using ultrapure water, and analyzed for size and zeta-potential.All size and zeta-potential analyses were performed using a nanosizer (Malvern Inst.) and a zeta-potential cell.The refractive index of water and protein solutions was 1.330

Rheological Analysis
Measurements of apparent viscosity were performed on the protein systems using a starch pasting cell attached to an AR2000ex controlled-stress rheometer (TA Instruments).Protein solutions with or without calcium were subjected to preshearing of 10 rad/s before the temperature was increased from 20 to 90°C at a shear rate of 15 rad/s for 15 min.This was followed by a peak hold step where the sample was held at 90°C at a constant shear rate of 15 rad/s for 300 s.N-Tetradecane was used as a solvent trap to stop evaporation of water from the samples.

Protein Profile Analysis
Reverse phase-HPLC was carried out on the protein systems using an Agilent 1200 HPLC with a poroshell C18 column for separation and a multi-wavelength detector at 214 nm (Agilent Technologies).The samples were prepared using 7 M urea denaturing buffer at pH 7.5 as described by Visser et al. (1991) before filtration using a 0.22-μm polyethersulfone (PES) filter (Carl Stuart Ltd.).The extent of denaturation of whey proteins in the samples after heating was assessed by adjusting to pH 4.7 using acetate buffer before centrifugation, as described by O'Kennedy and Mounsey (2009).
Size exclusion chromatography was performed on the supernatants from ultracentrifugation at 100,000 × g (1 h at 22°C) of the protein systems at pH 6.8 using TSK gel G2000SW XL and G3000SW XL columns (7.8 × 300 mm) arranged in series (Tosoh Bioscience).The HPLC modules were a 2695 Waters system with a 2847 dualwavelength detector with absorbance measurement at 214 and 280 nm.The samples were prepared to a target final protein concentration of 2 mg/mL before filtration through a 0.45-μm PES filter using the method previously described by Buggy et al. (2018).

Heat Stability
The heat stability of the protein samples prepared as described above was assessed using a temperaturecontrolled oil bath at either 120 or 140°C (Elbanton Instruments B.V.) according to the method of Davies and White (1966).The samples without mineral addition were analyzed at 140°C, and those with 1.75 mM added Ca were analyzed at 120°C to obtain measurable data for heat coagulation time at pH 6.8.

Statistical Analysis
All protein samples were prepared and analyzed in triplicate (n = 3), comparing the means using ANOVA and post hoc Tukey's test in Minitab 17 statistical software (Minitab).A P-value < 0.05 for a difference between treatment groups was considered significant.

Compositional Analysis
The protein contents of the final powders used to prepare the protein systems were as expected for whey concentrate and skim milk.The AWP had an average protein content of 84 ± 0.93% (wt/wt), and the SWP had an average protein content of 81.7 ± 1.37%.The SMP had an average protein content of 35.9 ± 1.06%.
For the protein systems measured by ICP-MS, no significant (P > 0.05) differences were observed in the concentrations of minerals between the unheated samples (SWCTRL, SWACTRL, AWCTRL, and AWACTRL) before ultracentrifugation (Table 1).For the samples that were ultracentrifuged, there was significantly (P < 0.05) less Mg in the ultracentrifugal serum of AWA85 compared with the serum phase of AWACTRL.There was significantly (P < 0.05) less Ca in the ultracentrifugal sera of AW85 and AWA85, respectively, compared with the sera of AWCTRL and AWACTRL, respectively.

Heat Stability
The heat stability of SWCTRL and SWACTRL without added Ca (Table 2) was significantly (P < 0.05) lower than that of the corresponding AWCTRL and AWACTRL.The heated samples SW85 and SW110 were significantly more heat-stable (P < 0.05) than SWCTRL.Likewise, the heat stability of SWA85 and SWA110 was significantly (P < 0.05) higher than that of SWACTRL.
Protein solutions prepared from sweet and acid wheys were sensitive to the addition of Ca at 7 mM, rapidly coagulating in the heat stability tubes on heating to 140°C.Consequently, the amount of Ca added and the oil bath temperature were reduced from 7 to 1.75 mM and from 140 to 120°C, respectively.The sensitivity of a model infant formula to Ca at pH 6.8 has been demonstrated previously (Joyce et al., 2017).The SW110 and the SWA110 samples (Table 3) displayed significantly (P < 0.05) higher heat stability compared with the corresponding unheated controls (SWCTRL and SWACTRL).The heat stability of the AW samples followed the same trend, but the differences observed upon heating were not significant.

A-C
Values within a row with different uppercase superscripts differ significantly (P < 0.05). 1 Values reported are mean ± SD of samples prepared from 3 different batches of milk. 2 SW = sweet whey; SWA = sweet whey adjusted; AW = acid whey; AWA = acid whey adjusted.Skim milk was added to attain a 60:40 whey: casein ratio or a 65:35 whey: casein ratio for adjusted samples.compared with the corresponding unheated control samples.The results also showed that after ultracentrifugation, there was a significantly (P < 0.05) higher proportion of aggregated (>300 kDa) material in the AWA110 samples than in the corresponding AWA85 samples.There was a significantly (P < 0.05) greater proportion of 15-30 kDa aggregate material in SWC-TRL, SWACTRL, AWCTRL, and AWCTRL than in the protein systems heated at either 85 or 110°C.The protein systems heated at 110°C had the lowest proportion of 15-30 kDa material.

Rheological Analysis
The protein systems prepared with 7 mM Ca had higher viscosity upon heating in the rheometer than Values within a column with different lowercase superscripts differ significantly (P < 0.05).A,B Values within a row with different uppercase superscripts differ significantly (P < 0.05). 1 Values reported are mean ± SD of samples prepared from 3 different batches of milk. 2 SW = sweet whey; SWA = sweet whey adjusted; AW = acid whey; AWA = acid whey adjusted.Skim milk was added to attain a 60:40 whey: casein ratio or a 65:35 whey: casein ratio for adjusted samples.the corresponding samples without Ca added (Figure 4).From the rheological data, the apparent viscosity of acid or sweet whey protein samples heated to 85 or 110°C without added Ca was the same as that of the corresponding unheated samples.Upon addition of Ca (Figure 4), viscosity increased for all samples, with coagulation commencing at 70°C on heating in the rheometer.

Particle Size Distribution and Zeta-Potential
In samples without added Ca, there were significant differences in particle size between the heated and unheated samples for all treatments (Table 4).The acid whey samples heated to 110°C had significantly (P < 0.05) larger particle sizes than their corresponding controls.However, there was no increase in the size of the particles in the protein solutions heated to 85°C compared with controls.
Adding 7 mM Ca to the unheated or heated protein systems significantly increased the particle size for some treatment groups.Sample AWCTRL with added Ca had a significantly (P < 0.05) higher particle size (202 nm) than the same sample without Ca addition (174 nm).Sample AW85 had a significantly (P < 0.05) larger particle size (266 nm) upon addition of Ca than the same sample without Ca addition (188 nm).Sample SW85 also had a significantly (P < 0.05) larger particle size (240 nm) upon addition of Ca compared with the same sample without Ca (187 nm).Conversely, the addition of Ca (7 mM) did not affect the particle size of any samples heated to 110°C except for AWA110, in which particle size was smaller.
When the heated and unheated protein systems without added Ca were analyzed for zeta-potential, no significant (P > 0.05) differences in charge were ob- profile of proteins and soluble protein aggregates in the supernatant fraction of protein systems without mineral addition after ultracentrifugation at 100,000 × g for 1 h at 22°C, as determined using size exclusion chromatography.SW = sweet whey; AW = acid whey; SWA = sweet whey adjusted; AWA = acid whey adjusted; CTRL, 85, and 100 denote no heat treatment and heating at 85°C and 110°C, respectively.Uppercase letters (A-C) denote significant differences (P < 0.05) for aggregated material (>300 kDa), in the same protein systems, compared for different heating regimes (e.g.SWCTRL, SW85, SW110).Lowercase letters (a-c) denote significant differences (P < 0.05) for aggregated material (15-30 kDa), in the same protein systems, compared for different heating regines (e.g., SWCTRL, SW85, SW110).Error bars represent SD. served, irrespective of the heating regimen (Figure 5) used.However, significant changes were observed with the addition of 7 mM Ca to the unheated or heated protein systems.Sample SWCTRL had a significantly (P < 0.05) lower zeta-potential (−21.3 mV) than the same sample without Ca addition (−28.7 mV).Likewise, AWCTRL had significantly (P < 0.05) lower zetapotential (−21.9 mV) upon addition of Ca than the corresponding sample without Ca (−27.1 mV).

DISCUSSION
This study compared the physicochemical characteristics upon heating of protein systems comprising whey (AW and SW; 60%) and casein (40%) mixed in proportions relevant to first-stage infant formulations.The role of CMP and its effect on whey protein functionality, specifically thermal stability, was investigated.Caseinomacropeptide is known to have excellent heat stability characteristics and lower Ca sensitivity than other proteins found in whey, particularly β-LG.It has been shown to increase the heat stability of dairy protein mixtures, where it constitutes up to 20% of total whey protein (Thomä et al., 2006).
The data in Table 1 indicate no significant differences between the levels of minerals found in the AW and SW protein systems.In commercially produced protein ingredients, the level of minerals should be higher in AW than in SW because the acidification step during manufacture removes minerals from the milk colloidal phase into the serum phase.In this study, ultrafiltration and diafiltration of the AW and SW retentates during subsequent processing removed small, soluble, unbound ions, including Ca, from the serum phase, reducing any differences in mineral content between AW and SW.
The data in Table 1 indicate that more Mg and Ca were bound to the large soluble protein aggregates and casein micelles in the heated AW and AWA protein samples than in their respective unheated controls, and this Ca and Mg is subsequently removed upon ultracentrifugation.This demonstrates that heating the protein systems results in loss of minerals from the serum to the colloidal phase, similar to the behavior in milk, as shown in previous studies (Wahlgren et al., 1990).The extent of mineral depletion from the serum was greater in AW systems than in SW systems.Values within a column and calcium treatment (all heat treatments within a Ca treatment group) with different lowercase superscripts differ significantly (P < 0.05).A,B Values within a column and heat treatment (both Ca treatments compared for each heat treatment) with different uppercase superscripts differ significantly (P < 0.05).

1
Values reported are mean ± SD of samples prepared from 3 different batches of milk.From the data in Table 2, we can see that the unheated sweet whey samples appeared less heat-stable than the acid whey samples upon heating in the oil bath.Significantly higher heat stability was recorded for the sweet whey samples after heating them to 85 or 110°C for 4 min, demonstrating the importance of sequential heating steps during the manufacture of protein ingredients.Although CMP is itself stable (Siegert et al., 2012), its effects on whey protein denaturation and aggregation are complex and largely pH-dependent (Martinez et al., 2010;Croguennec et al., 2014;Gaspard et al., 2020).From the heat stability data (Table 2), heating SW increased its heat stability on subsequent heating, which concurs with the findings of Croguennec et al. (2014), who reported that the presence of CMP reduced aggregate formation, hydrodynamic particle size, and turbidity of whey protein solutions at pH 6.7 when heated to 90°C.The differences observed in heat coagulation time between the unheated AW and SW protein systems may be related to differences in the serum mineral profiles of the 2 systems (Table 1), with AW systems displaying significantly lower Mg in the serum phase.Soluble cations in the serum phase are known to affect the heat-induced buffering capacity of protein systems during subsequent thermal processing.
Preheating all samples at pH 6.8 increased the complexation of whey protein with the micellar caseins.It also increased the formation of whey protein-whey protein and casein-whey protein aggregates in the serum phase compared with the respective control samples, as evidenced by increased particle size (Table 4).Beaulieu et al. (1999) reported that increasing the whey-tocasein ratio in model milk systems from 20:80 to 60:40 changed the profile of aggregates found after heating to 95°C for 5 min at pH 6.7, leading to the formation of large amounts of whey protein-whey protein aggregates.Studies have reported that the generation of small protein aggregates can improve the stability of whey protein beverages (Ryan and Foegeding, 2015) and model IF to subsequent heating by preventing aggregation during HTST treatment (Joyce et al., 2017).
Increasing the amount of protein in the samples affected the heat stability of SWACTRL only, lowering it by almost 50% compared with SWCTRL.This was probably a result of SWACTRL having more true whey protein present than SWCTRL.Several studies have concluded that increasing the whey protein concentration results in increased denaturation and aggregation on thermal processing (Fitzsimons et al., 2007;Buggy et al., 2018).The addition of Ca (Table 3) led to a considerable decrease in heat stability in all samples at pH 6.8 compared with samples without added Ca (Table 2).The preheated samples appeared to show an increase in heat stability compared with controls, which may be related both to the degradation of amino groups needed for cation-mediated crosslinking and to changes in serum ion levels.Calcium has been shown to destabilize bovine milk-based systems (McSweeney et al., 2004;Sievanen et al., 2008), with several studies demonstrating that adding Ca salts decreases electrostatic repulsion between proteins while increasing salt bridging and pH-dependent ionic effects on buffering capacity of serum from milk-based systems (Singh et al., 2007;On-Nom et al., 2012;Crowley et al., 2014).Whey proteins are known to aggregate in the presence of Ca at room temperature even without heating (i.e., cold gelation) via noncovalent interactions such as Van der Waals and hydrophobic interactions, as well as electrostatic interactions (Barbut and Foegeding, 1993;Bryant and McClements, 1998;Veerman et al., 2003).
The flow characteristics of the protein systems under the conditions studied (Figure 4) suggest there was little structure formation upon heating, with the samples showing lower viscosity with increasing temperature.Upon adding 7 mM Ca to the SW protein systems, the flow curve moved upward at close to 70°C, with the samples coagulating in the rheometer; a similar response was also observed for AW samples (results not shown).This temperature is greater than the onset of denaturation of the 2 main whey proteins, α-LA and β-LG.At temperatures >70°C, Ca-mediated and heatinduced aggregation behavior is accelerated.Murphy et al. (2014) reported that the loss of secondary structure of β-pleated sheet and α-helices in WPI-phosphocasein mixtures by Fourier transform infrared spectroscopy was associated with β-LG.These molecular changes are accompanied by exposure of the free thiol group in β-LG, which the authors associated with increases in viscosity seen in infant milk formula at temperatures >70°C.
The increases in particle size for the AW and SW protein systems observed at 110°C compared with controls were likely related to the interaction of whey proteins with casein micelles, although the extent of this at pH 6.8 might be as low as 30% (Anema and Li, 2003).Indeed, a significant amount of soluble whey protein-casein or whey protein-whey protein aggregates may have formed in the samples (Anema and Klostermeyer, 1997;Beaulieu et al., 1999;Vasbinder and De Kruif, 2003).As stated earlier, many of these aggregates can significantly affect thermal stability, especially on subsequent heating in the presence of Ca and other cations.
When 7 mM Ca was added to the protein solutions, there was a significant increase (P < 0.05) in the particle size of unheated protein systems and those heated to 85°C compared with those without added Ca.The protein systems heated to 110°C did not show any Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATION significant increases in particle size upon Ca addition compared with the corresponding samples without Ca addition.This may be due to a lack of binding sites for cross-linkage, which may be caused by conformational changes occurring in the protein structure or the direct effect of thermal degradation.Both the phosphoserine residues on the caseins and the carboxyl groups on the whey proteins form complexes with themselves and each other at high temperatures (Gaucheron, 2005;Singh et al., 2015;Villumsen et al., 2015b).The significantly (P < 0.05) higher level of denaturation (Figure 2) observed between the whey protein component of the protein solutions heated to 110°C and those at 85°C supports the fact that the samples heated to 110°C were more extensively modified.The effect of the calcium-mediated aggregation of milk proteins can clearly be seen in the particle size data even before secondary heating (Table 4); this, together with the zeta-potential data (Figure 5), may help explain some of the subsequent decreases observed in heat stability of the protein systems.The effect of adding 1.75 mM Ca on the heat stability of the protein systems would, however, be expected to have a lesser effect than addition of 7.0 mM Ca.
The protein profiles (Figure 3) showed a significant increase in higher MW material upon heating, attributable to whey protein-whey protein or whey protein-casein soluble aggregates in the serum.It is known that whey proteins, when heated, can both unfold and bind with each other or κ-casein to form soluble aggregates in the serum or bind directly with κ-casein to form aggregates on the surface of the casein micelle in a temperature-and pH-dependent manner (Corredig and Dalgleish, 1996;Anema and Klostermeyer, 1997;Donato and Dalgleish, 2006).The data presented in Figure 3 demonstrate that heating increased the proportion of the largest size (>300 kDa) aggregate material and decreased the proportion of smaller aggregates (15-30 kDa).The protein profiles concur with the particle size data, where heating increased the particle size of the protein systems.These data suggest that heating the protein systems resulted in the formation of aggregated material through mechanisms such as whey proteinwhey protein interactions or whey protein-casein interactions, similar to those found in both whey-dominant model milk systems and milk.
Heating the sweet whey samples to 110°C improved the heat stability of the protein systems upon subsequent heating.This agrees with previous studies, which found that heating whey proteins in the presence of CMP affected both aggregation and gelation, particularly at neutral pH, reducing the aggregate size (Croguennec et al., 2014) and affecting gelation properties (Veith and Reynolds, 2004).Several authors have demonstrated that reducing aggregate size is a key factor in stabilizing milk-based systems that are subjected to sequential heating steps (Ryan et al., 2012;Ryan and Foegeding, 2015;Joyce et al., 2017;Buggy et al., 2018).From the heat stability data, it was clear that preheating did not have the same stabilizing effect during secondary heating for the AW protein systems as for SW systems, which may reflect the presence of more favorable aggregate profiles.

CONCLUSIONS
In this study, we showed that preheating whey to achieve controlled aggregation of whey proteins improved its stability to subsequent heating.The implications of this work are 3-fold: (1) the source of the whey is a key determinant of functionality, with mineral-depleted acid whey ingredients without Ca addition having greater heat stability in resultant protein systems than sweet whey; (2) the presence of CMP in the sweet whey did not significantly improve the heat stability of the resultant protein systems at the levels studied; and (3) preheating of the sweet whey improved the heat stability of the resultant protein systems on subsequent heating.
Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATIONmanufacture to achieve desired thermal stability and subsequent reactivity when used in combination with skim milk and on fortification with minerals such as calcium during finished product processing.

Figure 1 .
Figure 1.Flow diagram of the preparation of protein systems containing either sweet or acid whey and skim milk with or without heating.SW = sweet whey; AW = acid whey; SM = skim milk; SWC = sweet whey concentrate; AWC = acid whey concentrate; SWP = sweet whey powder; AWP = acid whey powder; CMP = caseinomacropeptide; SMP = skim milk powder.
Figure3.Molecular weight (MW) profile of proteins and soluble protein aggregates in the supernatant fraction of protein systems without mineral addition after ultracentrifugation at 100,000 × g for 1 h at 22°C, as determined using size exclusion chromatography.SW = sweet whey; AW = acid whey; SWA = sweet whey adjusted; AWA = acid whey adjusted; CTRL, 85, and 100 denote no heat treatment and heating at 85°C and 110°C, respectively.Uppercase letters (A-C) denote significant differences (P < 0.05) for aggregated material (>300 kDa), in the same protein systems, compared for different heating regimes (e.g.SWCTRL, SW85, SW110).Lowercase letters (a-c) denote significant differences (P < 0.05) for aggregated material (15-30 kDa), in the same protein systems, compared for different heating regines (e.g., SWCTRL, SW85, SW110).Error bars represent SD.

Figure 4 .
Figure 4. Average apparent viscosity of the sweet whey protein system prepared at 110°C × 4 min (SW110) with and without 7 mM added calcium.
Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATION 2 SW = sweet whey; SWA = sweet whey adjusted; AW = acid whey; AWA = acid whey adjusted.Skim milk was added to attain a 60:40 whey: casein ratio or a 65:35 whey: casein ratio for adjusted samples.

Figure 5 .
Figure 5. Average zeta (Z)-potential of sweet or acid whey protein systems subjected to heating at 85 or 110°C for 4 min with or without subsequent addition of 7 mM calcium.
Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATIONand 1.450, respectively, with an absorbance of 0.001 used.Measurements were carried out at 25°C.

Table 1 .
Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATION Mineral composition of sweet whey, sweet whey adjusted, acid whey, and acid whey adjusted protein systems before and after ultracentrifugation at 100,000 × g; samples were analyzed in triplicate and presented as mean and SD

Table 2 .
Heat coagulation time (min) at 140°C for the different protein systems, unheated or heated, and without added calcium 1

Table 3 .
Corrigan et al.: WHEY SOURCE AND HEAT-INDUCED AGGREGATION Heat coagulation time (s) at 120°C for the different protein systems, unheated or heated, and with 1.75 mM added calcium 1

Table 4 .
Mean particle size (Z-average diameter; nm) of protein systems with and without added calcium 1