Effect of temperature and protein concentration on the protein types within the ultracentrifugation supernatant of liquid micellar casein concentrate

Liquid micellar casein concentrate (MCC) is an ideal milk-based protein ingredient for neutral-pH ready-to-drink beverages. The texture and mouthfeel of liquid MCC-based beverages depend on the beverage protein content, as well as the composition of soluble proteins in the aqueous phase around the casein micelle. The objective of this study was to determine the composition of soluble proteins in the aqueous phase around the casein micelles in skim milk and liquid MCC containing 7.0% and 11.6% protein content. Skim milk was pasteurized and concentrated to 7% protein content by microfiltration and then to 18% protein content by ultrafiltration. The 18% MCC was then serially diluted with distilled water to produce 11.6% and 7.0% protein MCC. Skim milk, 7.0% MCC, and 11.6% MCC representing starting materials with different protein concentrations were each ultracentrifuged at 100,605 × g for 2 h. The ultracentrifugation for each of the starting materials was performed at 3 different temperatures: 4°C, 20°C, and 37°C. The ultracentrifugation supernatants were collected to represent the aqueous phase around the casein micelle in MCC solutions. The supernatants were analyzed by Kjeldahl to determine the crude protein, casein, and casein as a percentage of crude protein content, and by sodium dodecyl sulfate PAGE to determine the composition of the individual proteins. Most of the proteins in MCC supernatant (about 45%) were casein proteolysis products. The remaining proteins in the MCC supernatant consisted of a combination of intact α S -, β, and κ-caseins (about 40%) and serum proteins (14–18%). Concentrations of α S -casein and β-casein in the supernatant increased with decreasing temperature, especially at higher protein concentrations. Temperature and interaction between temperature and protein explained about 80% of the variation in concentration of supernatant α S - and


INTRODUCTION
Micellar casein concentrate (MCC) is also known as microfiltered milk protein or micellar casein, and it is defined by the American Dairy Products Institute as concentrated milk protein produced by microfiltration (MF) to have a modified ratio of casein to whey protein, typically in the range of 82:18 to 95:5 (ADPI, 2021).Compared with milk protein concentrate, most of the milk-derived whey proteins (MDWP) in MCC have been removed by MF.The purity of MCC is defined as the amount of casein as a percentage of true protein (Whitt et al., 2022).Removal of 95% and 69% MDWP produces MCC with a purity of 94% and 90%, respectively (Whitt et al., 2022).Removal of MDWP prevents the production of sulfur-eggy off-flavors that originate from the thermal degradation of MDWP during high heat processing (Cheng et al., 2019;Jo et al., 2019;Vogel et al., 2021).The MCC is available as spray-dried powders or as fresh liquid concentrates.In liquid form, MCC is easier to reconstitute for beverage production and has a more neutral flavor, whereas spray-dried MCC has been described as having a cooked and corn chip-like flavor (Carter et al., 2018).Its neutral flavor and ease of reconstitution make liquid MCC an ideal milk-based protein ingredient for neutral-pH ready-todrink beverages.
Micellar casein concentrate is composed mainly of caseins in micellar form.Casein micelles of bovine milk have an average diameter of about 0.15 µm (Fox and Brodkorb, 2008), and they are retained in the retentate during MF of skim milk.Casein micelles are composed of α S1 -CN, α S2 -CN, β-CN, and κ-CN that are held together by hydrophobic bonds, electrostatic interactions, hydrogen bonds, and calcium phosphate nanoclusters (Lucey and Horne, 2018).The molar ratio of caseins α S1 :α S2 :β:κ inside the micelle is approximately 4:1:4:1 (Swaisgood, 2003).Multiple models for the structure of casein micelles have been proposed: submicelle model (Schmidt, 1982;Walstra, 1999), nanocluster model (Holt, 1992), and dual-binding model (Horne, 1998(Horne, , 2002)).Generally, most of the κ-CN have been suggested to reside on the surface of the casein micelle, and phosphorylated α S -CN and β-CN have been suggested to interact with calcium phosphate and other caseins to form the interior of the casein micelles (Dalgleish and Corredig, 2012).
The dissociation of casein micelles, especially at high protein content, has been linked to sharp viscosity increases and cold gelation in liquid MCC (Dunn et al., 2021).Ultracentrifugation of liquid MCC separates the colloidal phase (as centrifugally sedimented casein micelles) from the aqueous phase (as supernatant).The supernatant of the liquid MCC contains the soluble proteins that have dissociated out of the casein micelles into the aqueous phase surrounding micelles and the remaining milk serum proteins.According to Carter et al. (2021a), the viscosity, texture, and mouthfeel of MCC-based protein beverages largely depends on the protein content, but these sensory attributes can also be influenced by the composition of individual proteins in the MCC supernatant.Dunn et al. (2021) reported that MCC protein content and MCC supernatant casein content were positively associated with the apparent viscosity of the MCC solution.Cheng et al. (2018) also found that beverages with high casein content had similar temperature-dependent viscosity behavior, and Misawa et al. (2016) reported that relative viscosity, as well as mouthcoating and throat cling, increased with increasing casein as a percentage of true protein in MCC.
According to a Mintel survey, 58% of US consumers consider "high protein" to be an important product claim in the nutrition drink category (Mintel Group Ltd., 2022).A protein content of at least 4.2% (i.e., 10 g of protein) per 240-mL beverage serving is needed for an "excellent source" of protein claim (FDA, 2022).As the protein content of the liquid MCC increases, so does the soluble protein content in the MCC supernatant, especially at refrigerated temperatures (Dunn et al., 2021).The increase in MCC protein content can be expected to reduce the mobility of casein micelles and increase viscosity as the distance between casein micelles decreases and steric interference increases (Lu et al., 2015).
The composition of individual proteins in the MCC supernatant will depend on each protein's propensity to dissociate out of the casein micelles.A variety of factors including ionic strength, temperature, pH, and calcium activity would influence casein micelle dissociation and the types of proteins that partition between the colloidal and aqueous phase (Horne, 1998;Post et al., 2012).These factors influence the dissociation of each type of individual protein to different degrees.α S -Casein and β-CN are calcium sensitive and readily precipitate with calcium salt addition (Post et al., 2012).β-Casein is temperature sensitive and dissociates upon cooling (Rose, 1968).κ-Casein complexed with β-LG dissociates from casein micelles upon heating (Dumpler et al., 2017).
In addition to intact α S1 -CN, α S2 -CN, β-CN, and κ-CN, the MCC supernatant would also contain proteolysis products of these caseins and serum proteins (i.e., residual MDWP not removed by MF).The serum proteins are mainly β-LG and α-LA, but can also include trace amounts of immunoglobulins, lactoferrin, and BSA.The serum proteins reside in the aqueous phase outside the micelles, so their concentration in the MCC supernatant depends on the level of MDWP removal during MF.The concentration of casein proteolysis products (CNPP) will largely depend on the activity of proteolytic enzymes, especially native milk plasmin, and the concentration and type of intact caseins that dissociated out of the micelles (de Rham and Andrews, 1982;Bhatt et al., 2017).As intact caseins from the casein micelles are released into the aqueous phase, they become more accessible to enzymatic activity and susceptible to proteolysis by plasmin (Bhatt et al., 2017).Of the different types of caseins, β-CN tend to have greater susceptibility to plasmin (Andrews, 1983).
Desired beverage viscosity, texture, and mouthfeel at specific protein contents of milk-based beverages could be achieved by controlling casein micelle dissociation and the composition of the different individual protein types in the aqueous phase around the micelle (Carter et al., 2021a).The objective of our study was to determine the composition of soluble proteins in the aqueous phase around the casein micelles in skim milk and liquid MCC containing 7.0% and 11.6% protein content.

MATERIALS AND METHODS
No human or animal subjects were used, so this analysis did not require approval by an Institutional

Experimental Design
Skim milk was pasteurized and microfiltered to produce liquid MCC with 7% true protein (TP) content (Dunn et al., 2021).The MCC was concentrated to 18% TP via UF.The 18% MCC was serially diluted with distilled water to 11.6% and 7.0% protein MCC and a range of other protein concentrations for sections of this work reported elsewhere (Dunn et al., 2021).Skim milk, 7.0% MCC, and 11.6% MCC representing starting materials with different protein concentrations were ultracentrifuged at 100,605 × g for 2 h.The ultracentrifugation for each of the starting materials was performed at 3 different temperatures: 4°C, 20°C, and 37°C.The ultracentrifugation supernatants were collected, and the CP, casein, and casein as a percentage of CP (CN%CP) content of the supernatants were determined based on their Kjeldahl nitrogen distributions.The composition of the proteins in the supernatant were determined by SDS-PAGE.The entire experiment was replicated 2 times.

Production of Liquid MCC
Liquid MCC was produced according to Cheng et al. (2018) with modifications as described by Dunn et al. (2021).Skim milk (~350 kg) was pasteurized (72°C for 16 s) and filtered at 50°C through a ceramic MF system to produce liquid MCC with about 95% milk serum protein removal (6.85% TP) using a 3-stage, 3× MF process as described by Zulewska and Barbano (2014).The protein concentration of the liquid MCC was then increased to 18% by UF at 50°C using a Pellicon 2 plate ultrafiltration apparatus (Millipore Sigma, Burlington, MA) equipment with a 10 KDal polyether sulfone membrane.The UF-concentrated MCC was serially diluted to 11.61% and 6.96% MCC using distilled water as described by Dunn et al. (2021).These concentrations were obtained based on target protein content of 6.3% and 10.5%, which represents 15 g and 25 g of protein per 240 mL of beverage serving.Samples of these diluted MCC and pasteurized skim milk were collected for ultracentrifugation.Two replicates of these samples were produced starting with different batches of skim milk and used for ultracentrifugation and further analysis.Diluted MCC samples were preserved with thimerosal immediately after processing to prevent microbial growth during ultracentrifugation.One milliliter of 10% wt/vol aqueous thimerosal (Thermo Fisher Scientific) was added to every 1,000 g of diluted MCC.

Determination of Apparent Viscosities of Skim Milk and Liquid MCC
The apparent viscosities (AV) of skim milk (3.4% TP) at 4, 20, and 37°C ± 1°C were measured in duplicate using a rotational viscometer (LV-DV2T, Brookfield Engineering Laboratories Inc., Middleboro, MA) equipped with a jacketed cup-and-bob fixture (Enhanced UL Adapter, Brookfield Engineering Laboratories Inc.).The sample was first tempered at 50°C for 10 min to erase any thermal history experienced during refrigerated storage before being placed in a water bath set to the target measurement temperature to allow for 5 min of equilibration.Temperature was maintained by placing 16 mL of tempered sample into viscometer cup and circulating water at the target temperature through the jacket during measurement.Each sample measurement reflected the average viscosity obtained under 30 s of shear after a 20-s shear equilibration period at a constant shear rate of 36.69 s −1 .
The AV of the liquid MCC were estimated based on the liquid MCC protein concentration as measured by Kjeldahl total nitrogen (TN) using best-fit equations adapted from Dunn et al. (2021) for calculating the AV of liquid MCC at 4°C, 20°C, and 37°C.The best-fit equations by Dunn et al. (2021) were developed using measured AV of 6%, 8%, 10%, and 12% MCC (at constant temperatures of 4°C, 20°C, and 37°C) that were produced (i.e., serially diluted) from the same batch of 18% UF-concentrated MCC used to produce the 6.96% and 11.61% MCC in this study.The AV of the serially diluted 18% MCC were measured at 6%, 8%, 10%, and 12%, but not at 6.96% and 11.61%, so AV of the MCC at these specific concentrations needed to be calculated using the best-fit equations reported in Dunn et al. (2021).At 4°C, AV of liquid MCC was estimated using best-fit second-degree polynomial Equation 1, which was approximated using AV data of MCC from 6.54% to 10.66% protein.This range was used because it provided better quadratic fit, which can be expected to provide better AV prediction at 4°C for the 2 protein concentrations of interest, than the range between 6.54% and 13.21% (increased variabilities of AV at high protein concentrations and low temperatures).The AV of liquid MCC was estimated (at 20°C and 37°C) using linear best-fit Equations 2 and 3, respectively, which were approximated using AV data of MCC from 6.54% to 13.21% protein (where p = protein concentration).

Ultracentrifugation of Skim Milk and Liquid MCC to Collect Supernatants
Ultracentrifugation of skim milk, 6.96% MCC, and 11.61% MCC samples were performed as described by Dunn et al. (2021).Briefly, samples were centrifuged (Sorvall Lynx 6000 Ultracentrifuge System, Thermo Fischer Scientific) at 100,605 × g for 2 h each at 4°C, 20°C, and 37°C.Each centrifuge tube contained ~40 g of sample, and the top 15 mL of supernatant was carefully removed with a pipet from each tube for further analysis.The supernatants were used to represent the aqueous phase around the casein micelles.

Chemical Analysis for Protein Content
Starting skim milk and diluted MCC were analyzed in duplicate for CP by Kjeldahl TN (method number 990.20; AOAC International, 2019), NPN (method number 990.21; AOAC International, 2019), and noncasein nitrogen (NCN; method number 998.05; AOAC International, 2019).True protein was calculated as TN − NPN × 6.38, casein was calculated as (TN − NCN) × 6.38, and serum protein (SP) content was calculated as (NCN − NPN) × 6.38.The ultracentrifugation supernatants were also analyzed in duplicate by Kjeldahl TN and NCN as described above, and the NPN was considered negligible in the supernatant of MCC.Crude protein was calculated as TN × 6.38.The CN%CP of the supernatant was calculated as (CN/ CP) × 100%.The composition of raw and intermediate materials in the MF and UF processing runs were monitored using a mid-infrared spectrophotometer as described by Dunn et al. (2021).

SDS-PAGE Analysis of the Aqueous Phase Around the Casein Micelles
Supernatants were analyzed using SDS-PAGE to determine the composition of individual protein types including α S -CN, β-CN, κ-CN, CNPP, and SP in the supernatant protein.The SDS-PAGE analysis was performed using the procedure of Verdi et al. (1987) applied to MCC supernatant samples with adjustments on sample protein loading according to Carter et al. (2021b).Briefly, a 10% to 20% polyacrylamide gradient gel with a stacking gel at pH 6.8 and separating gel at pH 8.8 was used to determine the relative proportion of protein types in the ultracentrifugation supernatants.Samples (0.1 mL) were diluted with Laemmli sample buffer (0.9 mL) of pH 6.8.Protein loading of the samples onto the gel was calculated using Kjeldahl reference protein content data to achieve an optical density of the predominant protein band for each sample in the range of 1.0 to 1.4.Gels were run at constant current of 20 mA and 30 mA per gel, while the sample passed along the stacking gel and separating gel, respectively.Gels were fixed, stained, and destained before quantitative analysis.Gels were scanned with USB GS 800 Densitometer using Quantity 1 1-D Analysis software (Bio-Rad Laboratories Inc., Hercules, CA) to obtain a relative protein composition of sample with gel scanning as described by Carter et al. (2021b).The relative proportions of each protein type in the supernatant were multiplied by the CP content of the supernatant (as determined previously by Kjeldahl analysis) to obtain the concentration of each protein type in the supernatant.

Statistical Analysis
The general linear models (GLM) procedure of SAS (version 9.4, SAS Institute Inc., Cary, NC) was used to compare the means of the supernatant CP, CN, and CN%CP across the 3 temperature treatments (4°C, 20°C, and 37°C) within each protein level and determine if they are different (P < 0.05).
The GLM procedure of SAS was also used to determine the effects of the starting material protein concentration (skim milk with 3.56% protein, 6.96% MCC, and 11.61% MCC), temperature (4°C, 20°C, and 37°C), and replicate on CP, CN%CP, α S -CN concentration, and β-CN concentration of the supernatant.Protein concentration (Prot) and replicate (Rep) were treated as categorical variables.Temperature (Temp) was treated as a continuous variable and mean centered transformed by subtracting the mean Temp (rounded to the nearest integer) from each of the Temp treatment to obtain values of −16, 0, and 17.Using these mean-centered data avoids co-linearity effects on statistical analysis (Glantz and Slinker, 2001).All interactions of these variables were included in the model along with both the linear and quadratic effects of Temp.The interaction terms were: Prot × Rep, Temp × Prot, Temp × Rep, Temp × Temp, Temp × Temp × Prot, and Temp × Temp × Rep.If the F-value for the full model was significant (P < 0.05) and explained a large amount of variation (R 2 > 90%), the significance (P < 0.05) of each term were determined.The effects of the categorical variables and their interactions were tested for significance using Prot × Rep as the error term, whereas the effect of temperature (linear and quadratic) as well as its interactions with categorical variables were tested for significance using the full model error.Nonsignificant terms were removed in a stepwise process to obtain a reduced model with only significant terms.The type III sum of squares (SS) of each significant term and coefficient of determination of the reduced model were

Composition of Protein in the Aqueous Phase Around Casein Micelles
Skim milk and liquid MCC are colloidal dispersions consisting of a dispersed phase (i.e., the casein micelles), and a continuous phase (i.e., the aqueous phase surrounding the dispersed phase that consists of water, SP, mineral, and lactose; Walstra et al., 2006).The CP, CN, and CN%CP of the aqueous phase around the casein micelle determined based on Kjeldahl analysis of skim milk and MCC supernatants are shown in Table 1.The CP, CN, and CN%CP increased (P < 0.05) with increasing protein concentration of the starting material and decreasing temperature (interaction effect [P < 0.05] between these 2 factors; Dunn et al., 2021).These observed increases in protein and casein concentrations indicate a migration of caseins out of the casein micelles and into the continuous aqueous phase.
To analyze the composition of individual proteins in the aqueous phase surrounding the micelles, the supernatants were analyzed using SDS-PAGE.Individual proteins (α S -CN, β-CN, κ-CN, CNPP, and SP) as determined based on the SDS-PAGE relative quantity of proteins multiplied by Kjeldahl estimated CP are shown in Figure 1.There were effects (P < 0.05) of protein concentration, linear and quadratic effects of temperature, and a strong temperature by protein concentration interaction effect on the concentration of proteins (α S -CN, β-CN, κ-CN, and CNPP) in the supernatant (Tables 2 and 3).In skim milk, a majority (80% on average) of the supernatant proteins consist of SP, and β-CN makes up most of the remaining proteins especially at 4°C (Table 3).These observations are consistent with those reported by Davies and Law (1983) that found high levels of β-CN relative to other caseins in the aqueous phase in skim milk.
In MCC, a majority (close to half) of the supernatant proteins consisted of CNPP, and the remaining proteins were a combination of intact α S -CN, β-CN, and κ-CN (about 40%) as well a low amount of SP (14-18% on average; Table 3).The primary intact casein monomer present in the aqueous phase of MCC was β-CN (about 15% of the total protein in the aqueous phase; Table 3).Practically all of the SP should reside in the aqueous phase, and the low amount of SP in the MCC aqueous phase represents the remaining 5% of SP from the original skim milk after removal of SP by MF to produce liquid MCC.No effect of protein concentration in MCC or temperature on SP concentration in the supernatant was detected.The protein content of the aqueous phase was higher in MCC as there was a higher level of both CNPP and CN monomers than in the aqueous phase of skim milk (Figure 1).
A high level of CNPP was observed in the MCC aqueous phase (Figure 1).The CNPP would have been the result of native milk proteases such as plasmin, kallikrein, and cathepsins.Plasmin is the major native protease in milk (Bastian and Brown, 1996).It is a relatively heat stable enzyme that can remain active after HTST pasteurization (Dulley, 1972).Two factors may have contributed to increased plasmin activity that led to the elevated CNPP concentration observed in the MCC supernatants.The first factor is increased casein and reduced β-LG concentration (Metwalli et al., 1998).
Sulfhydryl containing compounds such as β-LG cause destabilization of plasmin beginning at 45°C, while casein protects plasmin from heat denaturation (Metwalli et al., 1998).The level of β-LG in MCC had been considerably decreased compared with in skim milk (95% milk-derived whey protein removal), such that most the proteins are caseins (93-94% casein purity).In addition to the increased availability of substrate for proteolysis, there would have been less inhibitory effect from β-LG (Politis et al., 1993) on plasmin activity because of removal of β-LG by MF and more protective effects from caseins on plasmin activity.The second factor is the processing temperature used in the production of MCC.At pH 7.4, plasmin showed optimal activity between 37°C (Bastian and Brown, 1996) and 45°C (Metwalli et al., 1998), and plasmin continued to show activity up to 65°C (Metwalli et al., 1998).The MCC was produced by the MF of skim milk at 50°C (ca. 4 h) on a pilot scale MF unit (Zulewska and Barbano, 2014), so plasmin may have been active and proteolyzing casein during MF processing.A combination of these factors may have resulted in the higher proportion of CNPP observed in the supernatant of MCC compared with skim milk.Elevated CNPP concentrations in the MCC aqueous phase could have consequences on sensory  and 37°C), protein concentration (skim milk, 6.96% MCC, 11.61% MCC), and interaction between temperature and protein concentration on the concentration of α S -CN, β-CN, κ-CN, and CNPP were significant (P < 0.05).Only the effect of protein level on the concentration of SP was significant (P < 0.05).properties through the generation of bitter peptides.
A measurement approach to efficiently quantify degree of proteolysis in MCC at industrial scale as well as further investigation on the sensory bitterness of MCC are needed.

Effect of Temperature and Protein Concentration on α S -CN and β-CN Concentration in the
Aqueous Phase Around Casein Micelles.Concentrations of α S -CN and β-CN in the aqueous phase around the casein micelle increased with decreasing temperature, especially at higher protein concentrations (P < 0.05, Figure 1).Temperature and interaction between temperature and protein explained a majority (about 80%) of the variation in concentration of α S -CN and β-CN (Table 2).Temperature × temperature, and temperature × temperature × protein interaction were also significant terms (P < 0.05) for the concentration of α S -CN and β-CN, although they explained a smaller proportion (9-14%) of the total variation.
Upon cooling from 37°C to 4°C, α S -CN and β-CN accounted for 29% and 28%, respectively, of the increase in protein concentration of 11.61% MCC supernatant, and 14% and 25%, respectively, of the increase in protein concentration of 6.96% MCC supernatant (Figure 1).These findings are consistent with a study on milk that showed β-CN and α S -CN contributed to the majority of protein increase in the supernatant upon cooling (Downey and Murphy, 1970).
Temperature had the strongest effect (explains 53% of total variation) on β-CN concentration (Table 2), which was an expected outcome as β-CN is the most hydrophobic milk protein (Farrell et al., 2004).This is consistent with previous studies that measured the change in β-CN composition in skim milk aqueous phase as temperature is decreased (Rose, 1968;Downey and Murphy, 1970;Davies and Law, 1983).As in these studies, β-CN was the primary contributor to the increase in casein content of skim milk and MCC supernatant as temperature is decreased to 4°C (Figure 1).
Temperature × protein interaction had the strongest effect (explains 45% of total variation) on α S -CN concentration.α S -Casein concentration had a marked temperature dependent increase at high protein concentration (11.61% protein MCC; Figure 1).Least squares means (LSM) of the α S -CN concentration in the aqueous phase around the casein micelle was higher (P < 0.05) in 11.61% MCC than in 6.96% MCC and skim milk (Table 3).In contrast, no difference (P > 0.05) was detected between the LSM of α S -CN concentration in 6.96% MCC and in skim milk (Table 3), further emphasizing the importance of the temperature × protein effect (P < 0.05).
Compared with α S -CN concentration in skim milk aqueous phase, α S -CN concentration in 11.61% MCC aqueous phase increased sharply with decreasing temperatures, and α S -CN behaved more similarly to β-CN in its dissociation out of the micelles (Figure 1).The increased migration of α S -CN out of the micelles in high protein (11.61%)MCC may be attributed to weakening electrostatic interactions of the α S -CN as the ionic strength of the aqueous phase around the casein micelles was decreased by MF.The ionic strength of the aqueous phase in MCC is weaker than in skim milk because MCC was produced by MF and UF, which removed most of the lactose and soluble minerals from the original skim milk.The MCC used in this study was produced by serial dilution with distilled water from 18% MCC, and it can be expected to contain a negligible amount of lactose and soluble minerals.Due to its inverse solubility property, decreasing temperatures will also cause a release of colloidal calcium phosphate into the aqueous phase (Barone et al., 2021).Among all the caseins, α S -CN is the most sensitive to changes in this mineral equilibrium as it contains the greatest number of phosphate centers (Holt, 2004).The degree of phosphorylation in caseins occurs in this order: α S2 -CN > α S1 -CN > β-CN > κ-CN, such that α S2 -CN would have the highest capacity for calcium binding (O'Mahony and Fox, 2013).At high protein concentrations, weaker ionic strength may lead to solubilization of colloidal calcium phosphate resulting in disruption of the salt bridges among caseins and dissociation of caseins into the aqueous phase around the micelles.In milk protein concentrate, Xu et al. (2016)  Means within the same row that do not share a common superscript are different (P < 0.05).
1 Crude protein content of the supernatant (g of protein/100 g of supernatant).Least squares means represent an average across 3 temperature treatments (4°C, 20°C, and 37°C) for each of the protein levels (skim milk, 6.96% micellar casein concentrate [MCC], and 11.61% MCC).Concentration was calculated based on the relative quantity of the individual protein as determined by SDS-PAGE multiplied by the CP concentration as determined by Kjeldahl total nitrogen.calcium phosphate that dissociates with caseins are released from the casein micelles as protein bound minerals, ions, or other mineral complexes.
Caseins are held together by both hydrophobic and electrostatic interactions (Walstra, 1990).Weakening hydrophobic interactions by decreasing temperature, especially in the case of β-CN, and disrupting electrostatic interactions by decreasing ionic strength, especially in the case of α S -CN, could result in dissociation of caseins out of the micelle.This was observed in the supernatants of MCC that showed increased casein content at lower temperatures and compared with skim milk.

Effect of Temperature and Protein Concentration on κ-CN, CNPP, and SP Concentration in the Aqueous Phase Around Casein Micelles.
Protein concentration explained the largest percentage (35.5%) of total variation in κ-CN concentration, followed by temperature (28.5%) and temperature × protein interaction (27.5%).Protein concentration explained a majority (62%) of total variation in CNPP concentration followed by temperature (23%) and temperature × protein interaction (15%).Protein was the only significant factor explaining changes in SP concentration in the aqueous phase around the micelle.The LSM of κ-CN, CNPP, and SP concentration in the aqueous phase around the micelle in each of the 3 protein levels (skim milk, 6.96% MCC, 11.61% MCC) were all significantly different (P < 0.05, Table 3), demonstrating the effect of protein concentration.
Protein concentration had a significant effect (P < 0.05) on concentration of κ-CN, CNPP, and SP.The increase in supernatant concentrations of individual proteins with increasing MCC protein concentration could be partly attributed to the increasing phase volume of the dispersed casein micelle phase (collected as precipitated pellet in ultracentrifugation).As the casein micelle phase volume increases with increasing protein content, the volume of the continuous phase (ultracentrifugation supernatant) decreases, resulting in a more concentrated supernatant.Casein micelles exhibit voluminosity and are not fully compressible, such that the volume occupied by hydrated casein micelles would be greater than the volume of the dried casein monomers that form it (de Kruif, 1998).As the volume is maintained and the protein content is increased, the volume of aqueous phase surrounding the casein micelles decreases, while the amount of soluble proteins remains the same.Therefore, an increase in supernatant protein concentration would be expected.
κ-Casein and CNPP were primarily affected by protein concentration, but their concentrations also showed temperature dependency.This suggests that κ-CN, which also has hydrophobic regions in its primary structure, underwent a degree of dissociation out of casein micelles due to temperature decrease.It also suggests that the soluble CNPP have hydrophobic properties and may be bitter peptides (Shinoda et al., 1985).This is consistent with results of Davies and Law (1983) that showed considerable increase in Γ-CN (proteolysis product of β-CN) in the supernatant of skim milk as temperature is decreased from 20°C to 4°C.The concentration of SP in the supernatant was affected only by protein concentration.This result was as expected because most of the SP would be expected to remain in the aqueous phase outside the micelles, and its concentration would only be influenced by the changes in the aqueous phase volume.Dunn et al. (2021) showed that apparent viscosity of MCC beverages increased with decreasing temperature and increasing protein concentration.Using best-fit equations provided by Dunn et al. (2021), the AV of 11.61% MCC was calculated to increase from 5 to 37 mPa when the temperature is decreased from 37°C to 4°C (Table 4).In comparison, skim milk AV was measured to only increase from 1 to 3 mPa when the temperature is decreased from 37°C to 4°C (Table 4).

AV of Skim Milk and Liquid MCC
Viscosity changes in MCC beverages follow a similar protein content and temperature dependency as α S -CN and β-CN concentration changes in the supernatant.As protein content increased and temperature decreased, α S -CN and β-CN concentration in the supernatant increased (i.e., caseins dissociate out of the micelles), and this would be accompanied by an increase in apparent viscosity.Using electron microscopy, Lu et al. (2016) showed that dissociated caseins form entangled networks in the continuous phase around the micelles, thus connecting and immobilizing the micelles.The steric hindrance and immobilization of the micelles through this entangled network formed from caseins that have dissociated out of the micelles could explain the AV increase that occurs when temperature is decreased and protein concentrations are high.

CONCLUSIONS
The MCC supernatant represents the aqueous phase around the casein micelle in MCC solutions.Most the proteins in MCC supernatant were CNPP, whereas about 40% of the proteins were a combination of intact α S -CN, β-CN, and κ-CN, and 14% to 18% were SP.Concentrations of α S -CN and β-CN in the supernatant increased (P < 0.05) with decreasing temperature, especially at higher protein concentrations.Temperature and interaction between temperature and protein explained about 80% of the variation in concentration of supernatant α S -CN and β-CN.Concentration of supernatant κ-CN, CNPP, and SP increased (P < 0.05) with increasing MCC protein concentration, and MCC protein concentration explained most, or all of the variation in supernatant κ-CN, CNPP, and SP concentrations.Predicted MCC apparent viscosity followed the same patterns of temperature and protein content dependent changes as in supernatant α S -CN and β-CN concentrations.Viscosity of MCC-based beverages could be achieved by controlling the composition of these supernatant proteins.
Pranata et al.:  PROTEIN COMPOSITION OF MICELLAR CASEIN CONCENTRATE SUPERNATANT calculated.The relative type III SS of each of the reduced models' terms were divided by the total type III SS of the reduced model and expressed as the percentage type III SS.The percentage type III SS serves as a measure of the relative effects of each term on the dependent variables analyzed and provides information on the percentage of variation in the model explained by each significant term.

Table 1 .
Pranata et al.:PROTEIN COMPOSITION OF MICELLAR CASEIN CONCENTRATE SUPERNATANT Crude protein 1 content, casein 2 content, and casein as a percentage of CP (CN%CP) 3 of ultracentrifugation supernatants of skim milk, 6.96% micellar casein concentrate (MCC), and 11.61% MCC (mean of 2 replicates) as determined by Kjeldahl analysis

Table 2 .
Percentage of type III sum of squares variation for concentration of α S -CN, β-CN, κ-CN, serum protein (SP), and casein proteolysis products (CNPP) in the supernatant as explained by model factors 1 1All model terms had a significant effect within column (P < 0.05) and NS indicates that means were not significantly different (P ≥ 0.05).2Prot = product type by protein levels (skim milk with 3.39% protein, 6.96% micellar casein concentrate [MCC], and 11.61% MCC); Rep = replicate; and Temp = ultracentrifugation temperature (4°C, 20°C, and 37°C).

Table 3 .
also observed increasing dissociation of caseins and colloidal calcium phosphate nanoclusters with increasing calcium removal by UF and diafiltration.It is unknown whether the colloidal Pranata et al.: PROTEIN COMPOSITION OF MICELLAR CASEIN CONCENTRATE SUPERNATANT Least squares means for concentration of α S -CN, β-CN, κ-CN, serum protein (SP), and casein proteolysis products (CNPP) in g/100 g of supernatant 1

Table 4 .
Pranata et al.: PROTEIN COMPOSITION OF MICELLAR CASEIN CONCENTRATE SUPERNATANT Apparent viscosities (AV; mPa•s) of skim milk at 3.4% protein and micellar casein concentrate (MCC) at 6.96% and 11.61% protein across 3 different temperatures (4°C, 20°C, and 37°C) 1 Means within the same row that do not share a common superscript are different (P < 0.05).AV of 6.96% and 11.61% MCC were estimated using best-fit equations adapted from Dunn et al. (2021) for predicting MCC AV from protein concentration at different temperatures.The AV of skim milk were mean of 2 replicates measured using a rotational viscometer. 1