Effects of spray drying and freeze drying on the protein profile of whey protein concentrate

Abstract

Whey protein concentrate (WPC) is consumed for its high protein content. The structure and biological functionality of whey proteins in WPC powders may be affected by the drying technique applied. However, the specific impact of spray drying and freeze drying on the overall protein profile of whey protein derived from sweet whey streams at scale is unknown. Herein, we examine the effects of commercial-scale freeze drying and spray drying on WPC to determine which method better preserves bioactive whey proteins, with the goal of helping the dairy industry create high value products that meet the growing consumer demand for functional dairy products. WPCs were produced from pasteurized liquid whey using either a commercial spray dryer or freeze dryer. A variety of analytical techniques, including enzyme-linked immunosorbent assay, polyacrylamide gel electrophoresis and bottom-up proteomics using liquid chromatography-tandem mass spectroscopy were used to identify, quantify and compare the retention of bioactive proteins in WPC before and after spray drying and freeze drying. In addition, the extent of denaturation was studied via solubility testing, differential scanning calorimetry and hydrophobicity assessment. There was little to no difference in the retention or denaturation of key bioactive proteins between spray-dried and freeze-dried WPC powders. There was a higher percentage of select Maillard modifications in freeze-dried and spray-dried powders than the control. The lack of significant differences between spray drying and freeze drying identified herein indicates that freeze drying does not meaningfully improve retention of bioactive proteins compared with spray drying when performed after multiple pasteurization steps.

1. Introduction

Whey proteins are used in a variety of applications, including sports nutrition, infant nutrition and other functional applications (e.g., foaming, gelation). In addition to being a complete source of amino acids, whey proteins have other bioactivities, such as antimicrobial, anticarcinogenic and immunomodulatory properties (Madureira et al., 2007). Whey is typically sold in dehydrated form, as, in this form, it does not require refrigeration due to its low moisture content and can be added to other food products on a dry weight basis (Jelen, 2011; Macwan et al., 2016). Whey protein concentrate (WPC) is a powdered form of whey that contains approximately 35–80% protein by weight, with the remainder consisting of carbohydrates (including lactose), fats, minerals and vitamins (Table 1) (Council, 2004; Morr & Ha, 1993).

Table 1:

Whey protein concentrate 80 (Hilmar 8000) typical macro and micro molecular composition.

Component	Approximate (%)
Protein (dry basis)	83.4
Lactose	5.5
Fat	4.5
Moisture	5.0
Ash	2.5

Spray drying is the most common method of dehydrating WPC in the dairy industry and involves the atomization of liquid whey into a hot air chamber to rapidly evaporate water, creating a porous powder product with a moisture content of 1–6% (Carter et al., 2018; U.S. Dairy Export Council, 2017). Advantages of spray drying include high throughput and cost-effectiveness (Singh Banjare et al., 2019; Wang et al., 2017; Wright et al., 2009). Typically, the inlet temperature (T_in) of a spray dryer is 160–255 °C, and the outlet temperature (T_out) is 60–120 °C (Padma Ishwarya & Anandharamakrishnan, 2017; Selvamuthukumaran, 2019). Because of the high thermal stress applied throughout the process, as well as air-interface stresses (e.g., hydrophobic unfolding) and shear stresses during atomization, spray drying can cause protein denaturation (Haque & Adhikari, 2015). The extent of denaturation can be influenced by holding times at these temperatures, as well as ionic or pH conditions (Bernal & Jelen, 1985; Bogahawaththa & Vasiljevic, 2020; Edwards & Jameson, 2014). Maintaining a low T_out can significantly reduce denaturation, whereas altering the T_in has minimal effect on protein denaturation because rapid water evaporation from atomized droplets cools the surrounding air and droplets, effectively preventing denaturation (Anandharamakrishnan et al., 2007; Oldfield et al., 2005). Additionally, particle residence times, or the amounts of time whey protein droplets spend in the drying chamber (typically 17–30 s), can be adjusted to reduce exposure to high temperatures (Haque & Adhikari, 2015; Schmitz-Schug et al., 2013; Singh & Van den Mooter, 2016). Other undesirable reactions, such as protein cross-linking, can occur during spray drying as well, decreasing both functional and nutritional properties (Meade et al., 2019). For example, high temperatures can induce the Maillard reactions between the reducing sugar lactose and amino acids. This reaction has been shown to block the free ε-amino group on lysine, which is crucial to human metabolism (Rudloff & Lönnerdal, 1992; Schmitz-Schug et al., 2013).

Though spray drying has many benefits such as high throughput, it may lead to denaturation (disruption of secondary, tertiary and quaternary structure) of key bioactive milk proteins, which can limit the product’s bioactive potential for consumers. Freeze drying (a.k.a., cryodesiccation, lyophilization) is an alternative method of drying food products that uses freezing and sublimation under vacuum rather than heat. Freeze drying includes three stages: freezing, sublimation and desorption. Though effective for heat-labile proteins like lactoferrin (LF), it has a higher cost and energy expenditure than spray drying, making it less desirable for high-volume dairy producers (Morel et al., 2022; Ratti, 2001). The economic impacts of commercial spray vs. freeze-drying whey have not been directly compared. However, comparisons have been made for other dehydrated food products where preservation of key nutrients is a concern, such as fruit powders. Barbosa et al. (2015) conducted a study on the drying of probiotic orange powder and found that freeze drying is approximately 6 times more expensive than spray drying per kilogram of water removed. Though freeze drying does not use hot air for dehydration, proteins can degrade due to stresses such as low temperature, freeze-thaw cycles and drying stresses (Roy & Gupta, 2004; Wang, 2000). For example, immunoglobulin G (IgG) has been shown to be very sensitive to freeze-drying conditions, with reported reductions of 25–50% after freeze drying and the formation of insoluble aggregates (Sarciaux et al., 1999; Elfstrand et al., 2022; Sotudeh et al., 2018). Additionally, more thermally stable proteins such as α-lactalbumin (α-La) and α-casein have been shown to undergo conformational changes due to dehydration stresses induced by freeze drying (Prestrelski et al., 1993).

The extent of denaturation and subsequent loss of bioactivity of whey proteins after freeze drying and spray drying has been found to differ in some previous studies. Vincenzetti et al. (2018) found that pilot-scale spray drying raw donkey milk significantly decreased lysozyme activity when compared with raw milk, whereas freeze drying did not have a significant impact. Similarly, Zou et al. (2022) found a significant fold change of peroxidase, serum albumin, lactadherin (LA), LF and lactoperoxidase (LPO; when measured via liquid chromatography-tandem mass spectroscopy, LC-MS/MS) in spray-dried raw camel milk compared to their control but did not see the same fold change for freeze-dried powders. In the same study, they noted an 86% decrease in LPO activity following spray drying of camel milk compared to a 50% reduction after freeze drying (Zou et al., 2022). Morel et al. (2022) observed that pilot-scale spray drying of purified LF resulted in a decrease in iron-binding capacity by 12–17% compared to pilot-scale freeze-dried LF. The authors also determined via differential scanning calorimetry (DSC) analysis that LF was 8–10% more denatured after spray drying than after freeze drying (Morel et al., 2022). Zhang et al. (2022) observed that the IgG content (measured via enzyme-linked immunosorbent assay, ELISA) of raw milk serum protein concentrate decreased following lab-scale spray drying yet remained unchanged after freeze drying. In the same study, the authors observed that freeze-dried powders exhibited a significantly higher retention of immunoglobulin A (IgA) and immunoglobulin M (IgM) compared with spray-dried powders (Zhang et al., 2022).

Some studies, however, indicate no significant differences in denaturation (and, therefore, retention) between spray-dried and freeze-dried whey protein powders. Although LF is very thermo-labile, previous studies have reported no significant differences in the retention or conformation of LF between spray-dried and freeze-dried whey protein powders (Wang et al. 2017; Zhang et al. 2022). Muuronen et al. (2021) found no differences in the retention of β-lactoglobulin (β-Lg), α-La, immunoglobulins, bovine serum albumin (BSA), LF and LPO (via high-performance liquid-chromatography. HPLC) between pilot-scale spray-dried and lab-scale freeze-dried native whey concentrate.

Though many of these studies have demonstrated the impact of freeze drying and spray drying at pilot scale, there is a need to understand the extent to which spray drying and freeze drying affect the retention and activity of milk proteins on a commercial scale. In this study, we examine the effects of commercial-scale freeze drying and spray drying on WPC to determine which method better preserves bioactive whey proteins, with the goal of helping the dairy industry improve product quality and meet the growing consumer demand for nutritious and functional dairy products. Herein, we assess the protein changes in WPC treated by spray drying and freeze drying using various analytical methods, including solubility, DSC, hydrophobicity assessment via anilinonaphthalene-1-sulfonic acid assay (ANS), LC-MS/MS based proteomics, polyacrylamide gel electrophoresis (PAGE) and ELISA.

2. Materials and Methods

2.1. Commercial Manufacturing of Powders

Three separate lots of raw milk were collected by Hilmar Cheese Company (Hilmar, CA, USA) from a combination of Jersey and Holstein cows and stored at 4 °C in bulk tanks (Figure 1). Each milk lot underwent high-temperature short-time (HTST) pasteurization at 72 °C for 16 s at a Hilmar cheese plant (Hilmar, CA, USA). Each pasteurized milk lot was made into commercial-scale American style cheeses using the renneting process, and the resulting sweet whey byproducts were collected. Each sweet whey underwent a sequence of purification steps, including fat separation, a second round of HTST pasteurization (72 °C for 16 s), ultrafiltration (10 kDa MWCO) and diafiltration, to concentrate the whey proteins. The pasteurized, ultra- and diafiltrated (20–35% solids) liquid feed (Feed-WPC) was provided as a control for each of the three lots (Figure 1). For each lot, a fraction was spray-dried using a multi-stage industrial-scale spray dryer (Custom Model, GEA, Düsseldorf, Germany) to make spray-dried WPC (SD-WPC; Fig 1). The flow rate of the incoming feed was 3,628–4,082 kg·h⁻¹, the T_in was 250–260 °C and the T_out was 70–82 °C. A separate fraction of each lot was freeze-dried using a pilot scale freeze dryer (Genesis SQ Super XL, SP Scientific, Warminster, PA USA) to make freeze-dried WPC (FD-WPC; Figure 1). The tray depth was 2 cm, the vacuum was held at < 13.3 Pa, and samples were maintained at –40 °C during drying for no more than 30 h. After sublimination, the final product temperature for the freeze-dried sample was 30 °C. All the above steps occurred at the Hilmar cheese plant (Hilmar, CA, USA).

Figure 1:

Schematic illustration of sample preparation of spray- and freeze-dried samples from three different lots.

2.2. Sample Preparation

The solubility method used to separate out soluble whey proteins from insoluble proteins, aggregates and insoluble caseins was adopted from Anandharamakrishnan et al. (2008), Grácia-Juliá et al. (2008) and Zhang et al. (2016), with some modifications, including solvent type, centrifuge speed and method of protein quantification. In this method, the pH of the whey protein solution is adjusted to 4.6 (the isoelectric point of caseins) and centrifuged to remove the insoluble precipitate that is formed. This method distinguishes native protein from aggregated proteins and caseins, as native proteins generally have hydrophilic surface properties that are preferential to intermolecular interactions with other proteins (Ryan et al., 2013). Ultrapure (18.2 MΩ−cm) water added to the powder (SD-WPC and FD-WPC) and liquid samples (Feed-WPC) to a final concentration of 1.0% w/w. Each sample was hydrated at 4 °C for approximately 24 h. Then each resulting solution was divided into two 25 g aliquots, and one aliquot was adjusted to pH 4.6 via addition of 0.25 M hydrochloric acid (0.15−1.0 mL; HCl, Millipore Sigma, Burlington, MA, USA), whereas the other was adjusted to the same final weight with ultrapure water. The pH 4.6 aliquot was centrifuged at 10,000 × g for 30 min at 21 °C (Eppendorf 5427 R Centrifuge, Hamburg, Germany) to remove any remaining caseins or insoluble proteins and aggregates. The supernatant was then extracted from the pellet (Figure 1).

The pH-adjusted and centrifuged sample will be referred to as the Soluble Fraction Sample (SFS), whereas the untreated sample will be referred to as the Total Sample (TS).

2.3. Solubility Assay

Protein quantification of each TS and SFS sample was carried out via a Pierce^™ BCA Protein Assay kit (Thermo Fisher Scientific, Waltham, MA, USA) to determine the amount of soluble protein in the SFS relative to the amount of protein in the TS. Prior to measurement, solutions of the SFS and TS were diluted 10-fold in ultrapure water. All samples were measured in triplicate, and the entire experimental sample preparation was repeated twice and then measured by BCA to examine variation in the assay. The percentage of total protein solubility was calculated using the following ratio: the amount of SFS to the amount of TS (Eq. 1).

$$TotalProteinSolubility(\text{\%})=\frac{Conc.ofpH4.6sample(SFS)}{Conc.ofuncentrifugedsample(TS)}$$ Eq 1.

2.4. Native and Sodium Dodecyl Sulfate (SDS) PAGE

Samples (both TS and SFS) were assessed via sodium dodecyl sulfate (SDS) PAGE using a Bio-Rad pre-cast Criterion^™ TGX^™ 4–20% polyacrylamide gel. All samples were diluted to approximately 2 μg ·μL⁻¹ in ultrapure water and prepared in a reduced and a non-reduced form. Reduced samples were prepared by adding 13.2 μL of Bio-Rad Laemmli loading buffer (62.5 mM Tris-HCl, pH 6.8, 10% glycerol, 1.0% lithium dodecyl sulfate, 5% Bromophenol Blue) and 5.5 μL of 1.0 M dithiothreitol (DTT) to 40 μL of the diluted samples. Samples were then heated to 95 °C for 5 min to fully denature the sample. Non-reduced samples were prepared in the same manner, except the reducing agent (DTT) was not added to the sample. Approximately 30 μg of protein was loaded into each well. An aliquot of Precision Plus Dual Color Standard^™ (Bio-Rad) was added to the first and last lane of the gel as a reference standard. The gel was run at 180 mV for 1 h in a Criterion Cell^™ with a running buffer of 1X Tris/Glycine/SDS (Bio-Rad). The gel was rinsed three times in ultrapure water for 15 min each, stained using Bio-Rad Coomassie Brilliant Blue G-250 for 1 h and destained in ultrapure water for 30 min.

SDS PAGE gels were imaged using a gel imager (LICOR Odyssey 9120 Near Infrared Gel Imager, Lincoln, NE, USA) at 700 nm.

2.5. Differential Scanning Calorimetry (DSC)

DSC measurements were performed as described previously by Anandharamakrishnan et al. (2007). Powder samples (FD-WPC and SD-WPC, TS only) were prepared in ultrapure water to 40% w/w, and approximately 18 mg of each sample was placed on an aluminum pan and hermetically sealed. The reference sample was an empty aluminum pan, and samples were measured using a DSC (Q20, TA Instruments, New Castle, DE, USA) from 20–100 °C at a rate of 5 °C ·min⁻¹. Each sample was measured in triplicate.

2.6. ANS Hydrophobicity Assay

ANS is a hydrophobic (apolar) fluorescent compound that binds to the nonpolar regions of proteins. The fluorescence quantum yield of ANS increases after binding to hydrophobic portions of proteins. A serial dilution of each sample type (TS only) was prepared in a standard curve ranging from 0.0625% to 1.50% (w/v) in 0.01 M HEPES buffer (4-[2-hydroxyethyl]-1-piperazineethanesulfonic acid, pH 7–7.5; Spectrum Chemical Mfg. Corp., New Brunswick, NJ, USA). A 2.5 μL aliquot of an 8 mM ANS (Spectrum Chemical Mfg. Corp.) in 0.01 M HEPES buffer was added to 500 μL of each sample. Approximately 200 μL of each sample (sample with ANS and blank samples without ANS) were aliquoted into wells of a 96-well fluorescent microplate. The plate was incubated for 15 min at 30 °C at 300 rpm. The relative fluorescent intensity (RFI) was measured with a spectrofluorometer (Molecular Devices SpectraMax M2 Plus Microplate Reader, San Jose, CA, USA) at excitation wavelength (λ ex) of 390 nm and emission wavelength (λ em) of 470 nm. The blank (including both protein and buffer) was subtracted from each measurement to get the final RFI value. The RFI was plotted against protein concentration, and the slope (S₀), or hydrophobicity index, of the line was calculated by linear regression analysis.

2.7. ELISA

The presence of bovine LF and IgG were analyzed (only TS samples) using ELISA. The kits used to detect LF and IgG were Bovine LF ELISA kit (ab274406, abcam, Waltham, MA, USA) and Bovine IgG ELISA kit (ab205078, abcam), respectively. Samples for the LF ELISA were diluted in assay buffer (provided in the kit) to obtain a range of 50,000–80,000 ng·mL⁻¹ of total protein. All samples for the LF ELISA were measured in quadruplicate. The samples for the IgG ELISA were diluted in assay buffer (provided in the kit) to obtain a range of 500–800 ng·mL⁻¹ total protein in assay buffer. All samples for the IgG ELISA were measured in triplicate. The standard provided in each ELISA kit was spiked at 50 ng·mL⁻¹ in each sample to determine the detection efficiency of the assay to account for any matrix effects. Equation 2 describes the detection efficiency.

$$DetectionEfficiency(\text{\%})=\frac{(AvgofSamplewithSpike-AvgofUnspikedSample)}{Spikeconcentration}$$ Eq. 2

All samples for the ELISAs were measured at 540 nm on a spectrophotometer (Molecular Devices SpectraMax M2 Microplate Reader). A quadratic fit for the standard curve was constructed for each assay, and the final concentration of each unknown was determined via SoftMax software (Molecular Devices SoftMax Pro Software, San Jose, CA, USA). The final values were normalized as a percentage against the total protein concentration that had been determined via BCA.

2.8. Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS)

Samples (both TS and SFS) were prepared according to the methods of Kim et al. (2023). The approximate starting concentration of each sample was 2.2 μg·μL⁻¹. All samples were prepared in duplicate. Chilled (–20 °C) ethanol (Sigma Aldrich, St. Louis, MO, USA) was added to each sample at a 5:1 ratio (by volume). Samples were then vortexed, chilled at –20 °C for 1 h and centrifuged at 12,000 × g for 20 min at 4 °C. The supernatant of each sample (containing peptides) was separated from the pellet (containing proteins). The pellets were then dried in a centrifugal evaporator for 70 min (max temp 37 °C) at the low boiling point setting (Genevac SP Scientific, Warminster, PA, USA) prior to reconstitution in 100 μL of 50 mM ammonium bicarbonate (Thermo Fisher Scientific) buffer with a pH between 7 and 8 for at least 24 h at 4 °C. A 2 μL aliquot of 550 mM DTT was added to each sample. Samples were incubated at 50 °C for 50 min at 300 rpm to reduce disulfide bonds. A 4 μL aliquot of 450 mM iodoacetamide was added to each sample. The samples were then incubated at room temperature for 60 min in the dark to alkylate thiol groups. An aliquot of 1 μg ·μL⁻¹ Trypsin/Lys-C (Promega, Trypsin/Lys-C Mix, Mass Spec Grade, Madison, WI, USA) was added to samples to achieve a 1:50 ratio of protease to protein. Samples were incubated at 37 °C for 16 h at 300 rpm to complete digestion.

Protein-derived tryptic peptides were extracted from the digested samples via C18 solid phase extraction (5 mL tube volume, 500 mg bed weight, 45 μm particle size, Sigma Aldrich). Cartridges were conditioned by adding 5 mL of ultrapure water, 5 mL of 80% acetonitrile (ACN; Merck, Darmstadt, Germany) with 0.1% trifluoroacetic acid (TFA, Merck, Darmstadt, Germany) and then another aliquot of 5 mL of ultrapure water. Samples were loaded by adding 1 mL of ultrapure water and then the 100 μL of sample. Samples were first washed with 5 mL of ultrapure water to remove any polar constituents, such as salts, then eluted with 5 mL 80% ACN with 0.1% trifluoracetic acid at a rate of one drop per second. The eluents were collected and dried in a centrifugal evaporator (Genevac SP; max temp 37 °C) using the HPLC setting.

The dried samples were reconstituted in 100 μL of 3.0% ACN and 0.1% formic acid (FA) for 24 h at 4 °C, before being further diluted by serially diluting 4.5 μL of the previous solution to 95.5 μL of 3.0% ACN and 0.1% FA (to achieve a target concentration of 100 ng·L⁻¹). The reconstituted samples were then injected into a Waters nanoACQUITY UPLC (Waters, Milford, MA, USA) with a C18 180 μm × 20 mm, 5 μm bead nano Acquity UPLC trap enrichment column (Waters, Milford, MA, USA) and a 100 μm × 100 mm, 1.7 μm bead Acquity UPLC Peptide BEH C18 analytical column (Waters). A 1 μL aliquot of each sample was loaded on the column and separated over the course of 60 min with a flow rate of 0.5 μL·min⁻¹ with the following gradient using solvent A (99.9 % ultrapure water with 0.1% formic acid) and solvent B (99.9% with 0.1 % formic acid): 3–11.5% B, 0 –10 min; 11.5–20% B, 10 –31 min; 20–30% B, 31–36 min; 30–95% B, 36 –45 min; 95% B, 45 –54.5 min; 95 to 3% B over 0.5 min. Then the column was re-equilibrated with 97% A for 5 min.

The sample was detected using an Orbitrap Fusion^™ Lumos^™ Tribrid^™ Mass Spectrometer (Thermo Fisher Scientific) with an electrospray voltage of 2,350 V and tube temperature of 300 °C. Full scan MS spectra were acquired in positive ionization mode over an m/z range of 300–2000 with a resolution of 60,000. The automatic gain control target was set to 4.0 × 10⁵, with a maximum injection time of 50 ms. The MS cycle time was set to 3 s. Following an MS scan, acquisition software automatically selected precursor compounds for MS/MS analysis based on the following criteria: ion-intensity threshold 5.0 × 10⁴, charge state 2–8 and exclusion time 60 s. Higher energy collision dissociation-based fragmentation was performed with 30% of collision energy. All MS/MS spectra were acquired in the positive ionization and auto scan range mode by the Orbitrap at a resolution of 60,000. The automatic gain control target was set to 5.0 × 10⁴ for all fragmentation techniques.

Two experimental replicates of each TS and SFS sample were analyzed. Data files were analyzed in Thermo Proteome Discoverer (v2.4.0.305) using the Sequest HT search engine against an in-house bovine milk protein database containing 496 protein sequences. Cleavage sites were set to C-terminal arginine (R) and lysine (K), and digestion specificity was selected as fully specific with a maximum of two missed cleavages. The precursor mass tolerance was set to 10 ppm with fragment mass tolerance of 0.1 Da. Minimum and maximum peptide length were set to 4 and 160 amino acids, respectively. Fixed modifications were set as carbamidomethylation (+57.021 Da) of cysteine. Dynamic modifications were set as methionine oxidation (+15.995 Da), serine and threonine phosphorylation (+79.966 Da) and N-terminus acetylation (+42.011 Da). Only proteins identified with high confidence (q-value <0.01) were retained for further analysis. Total protein abundances were derived from the sum of the intensities (area under the curve of the extracted ion chromatogram) of all peptides identified from each protein. Values for select proteins were reported as relative abundances against the total measured abundance of all detected proteins.

A subsequent database search for Maillard reaction products was conducted using the following additional dynamic modifications for all amino acids: carboxyethylation (CEL, +72.021 Da), hexose (1 x Hex, +162.053 Da), 2-hexose (2 x Hex, +324.106 Da) and pyrraline (PYR, +108.005 Da), based on previous methods (Renzone et al., 2015).

2.8. Statistics

Measurements from each assay for each sample type (Feed-, FD- and SD-WPC) were compared using a one-way analysis of variance (ANOVA) with post-hoc Tukey’s Honest Significant Different (HSD) tests (p-value < 0.05), while treatments (SFS vs. TS) within each sample type were compared via Student’s t-tests (p-value < 0.05). All statistical analyses were carried out in R statistical software (RStudio Version 2023.06.2+561, Posit Software, PBC Vienna, Austria).

3. Results and Discussion

3.1. Solubility Assay

The solubility assay carried out in this study serves as a reliable indicator for assessing the quantity of denatured proteins present in our samples. Native whey proteins remain soluble at pH 4.6, the isoelectric point of caseins. Therefore, if a whey protein is insoluble at pH 4.6, it could be due to aggregation or denaturation and, therefore, loss of native structure (De Wit & Van Kessel, 1996; Ryan et al., 2013). The protein solubility for all samples was on average 77–85% (Table 2). The solubility of the Feed (control) and FD-WPC were not significantly different from one another; however, the solubility of SD-WPC was significantly lower than that of the Feed and SD-WPC (Table 2; p < 0.05). The solubility values for SD-WPC identified herein (83 ± 4 %) with T_out 70–82 °C aligned with solubility values as reported by Tovar Jimenez et al. (2012), who detected 80–85% and 75–82% solubility for spray-dried powders produced with T_out of 60 °C and 100 °C, respectively. The significant difference observed in solubility for SD-WPC, but not FD-WPC (Table 2), may be attributed to the low solubility value of one lot (Supplementary Table S1). The solubility likely could have been increased by rehydrating the powder in an ionic solvent to induce the salting-in of proteins and disruption of hydrophobic interactions.

Table 2:

Average solubility (%)¹ of the control (Feed-WPC) freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC) (n=3). Average change in enthalpy (J×g⁻¹)¹ and peak temperature (°C)¹ of FD-WPC) and spray-dried whey protein concentrate SD-WPC (n=3 per sample type).

Measurements	Feed-WPC	FD-WPC	SD-WPC
Solubility (%)	81.88 ± 3.56 a	83.38 ± 4.00 a	77.36 ± 7.36 b
Enthalpy (J×g−1)	N/A	3.35 ± 0.264 a	2.96 ± 0.216 b
Peak Temperature (°C)	N/A	72.80 ± 0.514 b	73.57 ± 0.356 a

¹Averages with differing superscripts (a,b) do not share significance (Tukey’s HSD, p < 0.05). Values are expressed as the average ± standard deviation.

3.2. SDS-PAGE

Gel electrophoresis, specifically SDS-PAGE, was used to identify differences in the protein profile among Feed, FD- and SD-WPC (Figure 2). Samples were prepared in both reducing (addition of DTT) and non-reducing conditions. The addition of the reducing agent DTT breaks disulfide bonds within and between protein molecules. This reduction results in the separation of protein subunits and can allow a better resolution of individual protein components. Non-reducing SDS-PAGE, on the other hand, maintains these disulfide bonds, keeping the protein subunits linked together. Reducing and non-reducing SDS-PAGE often have differing band patterns.

Figure 2:

SDS-PAGE (reducing and non-reducing) of the control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC) of three separate production lots. (a) Total Samples (TS) and (b) Soluble Fraction Sample (SFS). Red lettering indicates the samples reduced by dithiothreitol activation, and black is non-reduced samples. Each color represents a different lot. “1” represents Feed-WPC, “2” represents SD-WPC and “3” represents FD-WPC.

The reduced samples had a band at 150 kDa, which corresponds to the reduced form of xanthine dehydrogenase oxidase (XDO), as indicated previously (Beyaztaş & Arslan, 2015). This band appeared as the same intensity across sample types but was not present in the non-reduced form of the samples. The reduced form of IgG appeared as two separate bands at ~50 kDa and ~25 kDa, which corresponds to heavy chain and light chain IgG, respectively (Ng, et al., 2010). IgG in the non-reduced form in all samples appeared just above 150 kDa for purified samples, but, in our gel, only appeared very faintly in the SFS gel (Figure 2b). LF appeared at ~80 kDa in the reduced form, but, in the non-reduced form, appeared as a faint band just below ~75 kDa (Figure 2a-b). Other bands correspond to BSA at ~60 kDa (both reduced and non-reduced form), caseins at ~20–25 kDa (both reduced and non-reduced form), β-Lg at ~18 kDa (both reduced and non-reduced form) and α-La at ~14 kDa (both reduced and non-reduced form; Figure 2a-b; Supplementary Figure S1a-b).

Overall, band intensities were similar across all samples (both reduced and non-reduced) for all identified whey proteins (Figure 2a-b). This observation aligns with the results of Zhang et al. (2022), who found minor differences in retention (measured via SDS-PAGE) of β-Lg, α-La, BSA, LF and LPO in spray-dried native whey protein concentrate compared with their spray-dried and freeze-dried protein concentrates. Additionally, our results align with those of Oldfield et al. (2005) who found minor differences in retention (measured via SDS-PAGE) of β-Lg, α-La, BSA, LF and LPO in spray-dried native whey protein concentrate (compared to their skim milk control).

For non-reduced SDS-PAGE samples, there was a band present at > 250 kDa among all TS-WPC samples (Figure 2a). This band was not observed in the SFS-WPC samples (Figure 2b). The absence of a band at the top of the gel in the SFS samples for SDS-PAGE (Figure 2b) indicates proteins or aggregates >250 kDa were excluded. Additionally, the band intensity at the top of the gel is greater in non-reduced samples than reduced samples, suggesting it may a be disulfide bonded aggregate (Figure 2a). In their respective studies, Zhang et al. (2022) and Singh & Amamcharla (2021) observed a >250 kDa band in whey protein samples analyzed via non-reducing SDS-PAGE. The presence of this band in all samples likely results from aggregate formation between milk proteins (Singh & Amamcharla, 2021; Zhang et al., 2022) prior to the evaporation treatments tested herein (e.g., due to pasteurization or cheesemaking). Future work could involve the extraction and digestion of the protein band above 250 kDa to determine its composition.

3.3. DSC

Higher change in enthalpy (ΔH) values correlate with a higher degree of proteins in their native (folded) state, whereas peak temperature indicates the temperature at which significant structural changes occur, such as unfolding or denaturation. There was a significant difference in the ΔH values and peak temperature (otherwise known as average melting temperature) between SD-WPC and FD-WPC (Table 2; p < 0.05). A possible explanation for the higher peak temperature in SD-WPC may be the presence of a greater abundance of thermally stable proteins (e.g., β-Lg and α-La) with higher denaturation temperatures in the SD-WPC sample. The observed peak temperatures for all samples were approximately 72.8–73.6 °C (Table 2; Supplementary Table S4), which was slightly lower than the range reported by Anandharamakrishnan et al. (2007) for spray-dried WPI powders produced within a T_out range of 60-120 °C. These differences in peak temperatures may be attributed to the differences in protein purity and composition of WPI (~90% protein w/w) vs. WPC (~80% protein w/w).

DSC has previously been used to understand the impact of spray drying on the denaturation whey proteins (Anandharamakrishnan et al., 2007; Haque, 2015) and compare the effects of freeze drying and spray drying on individual whey protein components, such as LF (Morel et al., 2022; Wang et al., 2017). However, it has not been used to compare the bulk denaturation profile of commercial spray-dried and freeze-dried WPC powders. The samples in our study are a heterogenous mixture of different proteins with differing denaturation temperatures, and it is likely the DSC results could be attributed to proteins that constitute the majority of WPC, such as β-Lg (50–60% w/w) and α-La (12–16% w/w). These protein constituents are more heat-stable than proteins such as LF and immunoglobulins, which constitute less than 10% w/w (Burrington & Agrawal, 2012). Therefore, these results reflect the average bulk denaturation profile with less impact from lower abundant individual components.

3.4. ANS assay

Each linear regression model was constructed from the averages of each dilution across three lots for each sample type (Figure 3). The average of the hydrophobicity index (S₀) of each lot was calculated from the slope. The higher the hydrophobicity index, the greater the surface hydrophobicity of proteins in a sample, which can indicate increased denaturation (Latypov et al., 2008). The hydrophobicity index was not significantly different among Feed-WPC, SD-WPC and FD-WPC (Figure 3; Supplementary Table S3; p > 0.05), which indicates there is no detectable difference in the average three-dimensional structure of proteins in these samples via this method (Cardamone & Puri, 1992). Though ANS florescence has previously been used to assess the surface hydrophobicity of heated solutions of spray-dried WPC powders by Moro et al. (2001), it has not been used to compare freeze-dried and spray-dried powders. The hydrophobicity index was numerically but non-significantly lower for SD-WPC than Feed-WPC and FD-WPC (Figure 3). If spray drying increased denaturation and the denatured proteins remained soluble, we would expect to see an increased in hydrophobicity index. The observed non-significant decrease in hydrophobicity index could be due to spray drying or resuspension-induced protein aggregation, which would decrease the available hydrophobic protein sites in solution. Our finding that the SD-WPC protein solubility was significantly lower than Feed-WPC could also suggest an increase in protein aggregation due to spray drying or post-spray drying resuspension.

Figure 3:

Relative fluorescent intensity (RFI) derived from ANS of the control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC) vs. sample concentration. Each linear regression line to determine the hydrophobicity index (S₀) was based on the average of triplicate measurements (n=3) for each dilution value of each sample type. Error bars represent standard deviations.

3.5. ELISA

The LF and IgG values were calculated as a percentage of the total protein, which was determined using the BCA assay (Figure 4a-b). Average values for LF were approximately 0.14–0.18% of total protein across sample types (Figure 4a; Supplementary Table S4). ELISA results indicated that LF concentration (expressed as a percentage of total protein) was significantly higher in Feed-WPC than both FD and SD-WPC, but that FD and SD-WPC did not differ from each other (Figure 4a).

Figure 4:

ELISA results as normalized percentage values (expressed as a percentage of total protein) for select proteins of the control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC). (a) Average normalized lactoferrin (LF) (%) and (b) Average normalized IgG (%) by sample type and treatment. Each bar represents the mean value, and the error bars indicate the standard deviation of triplicate measurements (n=3) for each sample. Differing letters (a.b) represent significant differences. (Tukey’s HSD, p < 0.05).

LF was specifically analyzed via ELISA due to its high bioactivity. LF has antibacterial, antioxidant and immunomodulatory properties, and is very heat labile compared to other whey protein constituents, with irreversible denaturation occurring at temperatures as low as 72 °C (Goulding et al., 2021; Li et al., 2021; Superti, 2020). Our results are consistent with findings of Zhang et al. (2022), who observed a significant decrease of LF (via ELISA) from their control (raw milk serum protein concentrate) to their spray-dried serum powder. Our results do not align with those of Morel et al. (2022), who found (via DSC) that spray-dried LF was more denatured than freeze-dried LF. However, the samples in their study were extracted from singly pasteurized skim milk and purified to approximately 95% prior to evaporation (Morel et al., 2022), whereas our samples were HTST pasteurized and dried as a bulk product with other protein constituents.

IgG was also assessed via ELISA, as it is the most abundant of the bovine Ig and can be affected by thermal treatments, with reported reductions of up to 25% post HTST pasteurization (Kaplan et al., 2022; Kummer et al., 1992). Average values for IgG were approximately 2–3% of total protein across all samples (Figure 4b; Supplementary Table S4). ELISA results revealed that the percentage of IgG was significantly higher in the Feed-WPC than in the FD-WPC. However, neither Feed-WPC nor FD-WPC differed from SD-WPC (Figure 4b). Our results partially do not with those by Zhang et al. (2022), who found significantly more IgG in their control (milk serum concentrate) than in the spray-dried raw milk colostrum powders, but no difference between their control and freeze-dried powders. However, in another study, Sotudeh et al. (2018) reported a reduction in IgG values in their spray-dried and freeze-dried bovine colostrum samples from their raw colostrum control, which aligns with our findings.

Overall, our ELISA results show no significant differences between freeze-drying and spray-drying in the preservation of LF and IgG. Our ELISA results for LF and IgG showed a reduction from Feed-WPC to dried samples (FD-WPC and SD-WPC). The SDS-PAGE results, however, did not indicate a loss of band intensity for these proteins across sample types (in either reduced or non-reduced samples, Figure 2a-b). SDS-PAGE, while useful for comparing overall protein composition and assessing aggregation, is not sensitive enough to detect small differences between samples. ELISA, on the other hand, is highly sensitive to select proteins in their native (intact) form and provides precise qualitative analysis. It is possible that both freeze- and spray-drying induce small amounts of degradation in LF and IgG, and these changes are not detected via SDS-PAGE.

3.6. LC-MS/MS

The Venn diagram (Figure 5a-b) displays the overlap of identified proteins via LC-MS/MS of the TS and P samples. The shared percentage of identified proteins in the TS was approximately 78% in the Feed-WPC sample, 85% in the FD-WPC sample and 87% in the SD-WPC sample (Figure 5a). In the SFS treatment group, the shared percentage of identified proteins was approximately 89% in the Feed-WPC sample, 88% in the FD-WPC sample and 85% in the SD-WPC sample (Figure 5b). For TS, FD-WPC shared 91% of the proteins identified in Feed-WPC, whereas SD-WPC shared 85% of the proteins identified in Feed-WPC (Figure 5a). Both SD-WPC and FD-WPC shared 89% of the proteins identified in Feed-WPC for the SFS group (Figure 5a).

Figure 5:

Venn diagrams of shared proteins among sample treatment of control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC). (a) Total Sample (TS) and (b) Soluble Fraction Sample (SFS), n=3 per sample type.

The distribution graph (Figure 6) illustrates the average abundance of the top 15 most abundant proteins across sample sets (TS and SFS). In general, the distribution of protein profiles across treatment types (TS and SFS) is consistent among sample types (Feed, SD, FD-WPC), (Figure 6a-b). Comparisons between specific proteins are discussed later.

Figure 6:

Distribution (as a percentage of total ion abundance) of control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC). (a) Total Sample (TS) and (b) Soluble Fraction Sample (SFS), n=3 per sample type. β-Lg, β-lactoglobulin; α-La, α-lactalbumin; GLYCAM-1, glycosylation-dependent cell adhesion molecule-1; LF, lactoferrin; BTN1A1, butyrophilin subfamily 1 member A1; BSA, bovine serum albumin; XDO, xanthine dehydrogenase oxidase; LA, lactadherin, H-FABP, fatty acid binding protein heart; α_s2-CSN, α_s2-casein; OPN, osteopontin; β-CSN, β-casein; and RAB3A, ras-related protein rab 3A.

The relative percentages of the whey proteins, α-La, BSA, LF, LPO and OPN (Figure 7a-f), as well the milk fat globular membrane (MFGM) proteins glycosylation-dependent cell adhesion molecule-1 (GLYCAM-1), XDO, butyrophilin subfamily 1 member A1 (BTN1A1), LA and the fatty-acid binding protein, heart (H-FABP; Figure 8a-e) were compared between sample type (Feed, SD-WPC, FD-WPC) and group (TS, SFS). Other proteins identified in the analysis of the top 10 abundances (α_s2-casein, β-casein, RAB3A) were compared as well (Supplementary Table S8). In the TS group, LF and LPO exhibited significantly higher relative percentages in the Feed-WPC (control) compared to SD and FD-WPC (Figure 7d, e), whereas α-La, BSA, OPN and GLYCAM-1 were significantly lower (Figure 7b, c, f; Figure 8a). In the SFS group, the relative percentage of BTNIAI was significantly lower in the Feed-WPC compared to SD and FD-WPC (Figure 8b), and GLYCAM-1 was significantly lower in SD-WPC compared to FD-WPC (Figure 8a). All proteins, except for β-Lg (Figure 7a), had significant differences in relative percentage between the TS and SFS between individual sample type. β-Lg, XDO, LA and FABP-H proteins had no significant differences between Feed-WPC, FD-WPC and SD-WPC in either the TS or SFS samples (Figure 7a; Figure 8d-e).

Figure 7:

Relative percentages of select proteins for pre-evaporated control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC). (a) Average relative percentage of β-lactoglobulin (β -Lg), (b) α-lactalbumin (α-La) (c) Average relative percentage of bovine serum albumin (BSA) (d) Average relative percentage of lactoferrin (LF) (e) Average relative percentage of lactoperoxidase (LPO) (f) Average relative percentage of osteopontin (OPN). Each bar represents the mean value, and the error bars indicate the standard deviation of triplicate measurements (n=3) for each sample. Differing uppercase letters (A, B, C) represent significant differences within the soluble fraction samples (TS) group. Differing lowercase letters (a, b, c) represent significant differences within the total sample (SFS) group (Tukey’s HSD, p < 0.05).

Samples with an asterisk represent differences between sample treatments (SFS vs. TS) between the same sample type (Feed, SD-WPC, FD-WPC) (Student’s t-test, p< 0.05).

Figure 8:

Relative percentage of select proteins for pre-evaporated control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC). (a) Average relative percentage of glycosylation-dependent cell adhesion molecule-1 (GLYCAM-1) (b) Average relative percentage of xanthine dehydrogenase oxidase (XDO) (c) Average relative percentage of butyrophilin subfamily-1 member-A1 (BTN1A1) (d) Average relative percentage of lactadherin (LA) (e) Average relative percentage of fatty-acid binding heart protein (FABP-H). Each bar represents the mean value, and the error bars indicate the standard deviation of triplicate measurements (n=3) for each sample. Differing uppercase letters (A and B) represent significant differences within the soluble fraction samples (TS) group. Differing lowercase letters (a and b) represent significant differences within the total sample (SFS) group (Tukey’s HSD, p < 0.05). Samples with an asterisk represent differences between sample treatments (SFS vs. TS) between the same sample type (Feed-WPC, FD-WPC or DD-WPC) (Student’s t-test, p< 0.05).

LC-MS/MS results revealed there were not any significant differences in retention of bioactive proteins between SD-WPC and FD-WPC (Figure 7a-f; Figure 8b-e), apart from GLYCAM-1 (Figure 8a). These findings align with the results of Muuronen et al. (2021), who found no differences in retention of the ratio of SD and FD native whey concentrate (via HPLC) of β-Lg, α-La, BSA, LF and LPO. Overall, β-Lg represented most of the protein distribution among all three treatment groups, accounting for approximately 76–77% of SFS (Figure 7a) and 78–82% of TS (Figure 8a). There was a significant decrease in the relative percentage of LF across sample types from TS to SFS (Figure 7d), which aligns with our SDS-PAGE results (Figure 2a-b). One possibility is that LF could be interacting with another protein (or self-aggregating) via disulfide bonds and is, therefore, being extracted during the pH adjustment and centrifugation steps during the SFS treatment. This hypothesis is supported by Xiong et al. (2020), who observed disulfide-induced β-Lg/LF interactions at a temperature range of 70–75 °C, confirmed via SDS-PAGE and LC-MS/MS. The double HTST pasteurization prior to evaporation may cause these interactions to occur.

OPN, a highly bioactive glycoprotein, was detected at approximately 0.1–0.3% in the samples (Figure 7f, Supplementary Table S8). There was a significantly higher retention of OPN in the FD and SD-WPC than Feed-WPC for the TS treatment group (Figure 7f). This finding partially agrees with the studies by Zhang et al. (2016) who observed no significant reduction in OPN (measured via LC-MS/MS) in spray-dried bovine milk powders compared to their raw milk control (Zhang et al., 2016). Similarly, Zou et al. noted an increased fold change of OPN in their spray-dried camel milk powders compared to the raw milk control, attributing it to OPN’s lack of the amino acid cysteine, which, they believed, prevented disulfide bond formation under heat treatment. (Zou et al., 2022). α-La was reported at 4–8% in the TS group and 12–13% in the SFS group (Figure 7b, Supplementary Table S7, S8). Anandharamakrishnan et al. (2008) observed that α-La was more soluble than β-Lg after spray drying at T_out ranging from 60–100 °C. α-La’s ability to bind to Ca²⁺ ions, thereby preventing self-aggregation and increasing its overall stability during heat treatment (Permyakov, 2020), could explain its increase in proportion from the TS to the SFS in our study. GLYCAM-1 (otherwise known as lactophorin) is a glycoprotein responsible for ligand binding during lactation (Dowbenko et al., 1993). The relative percentage of GLYCAM-1 was significantly higher in FD-WPC powders compared to Feed-WPC in both TS and SFS groups (Figure 8a, Supplementary Table S7, S8). Additionally, it was significantly higher in FD-WPC compared to SD-WPC for the SFS group (Figure 8a, Supplementary Tables S7, S8). These results partially align with those of Zou et al. (2022), who saw a significant increase of GLYCAM-1 in spray-dried camel milk powders compared to their raw milk control but did not see this increase in their freeze-dried powders. It is possible that the increase of abundance values from for these proteins is due to the decrease in abundance of major thermo-labile proteins (e.g., LF, LPO and XDO/XO). This decrease could have led to an increase in ionization efficiency and detection of more heat stable proteins (Liu, Zhang, et al., 2020; Page et al., 2007; Wei et al., 2022).

The remaining proteins (including the MFGM proteins XDO, BTN1A-1, LA and FABP-H) had a lower relative percentage in the TS than the SFS, suggesting they were extracted during the soluble fraction procedure (Figure 7d, e; Figure 8b-e). XDO is the dehydrogenated form of the enzymatic MFGM protein xanthine oxidase (XO), which catalyzes the reaction of hypoxanthine to xanthine and, subsequently, uric acid (Enroth et al., 2000; Ozturk et al., 2019). The relative percentage of XDO was significantly higher across sample types in the TS (1–1.4%) than SFS (0.03–0.06%; Figure 8b), which was also confirmed by the SDS-PAGE (Figure 2a-b). Additionally, XDO may be a component of the non-reduced aggregates present at the top of the SDS-PAGE gel (Figure 2a-b), as it only appears in the reduced samples (Supplementary Materials, Figure S1b). XDO relative percentage did not differ across Feed-WPC, FD-WPC and SC-WPC (Figure 8b). This finding does not align with that of Zhang et al.’s 2016 study, which found a complete loss of XDO in the soluble protein fraction after spray drying via proteomics. The difference may arise from the sample differences between these studies (a pasteurized and then spray dried milk vs. whey protein derived from cheesemaking with two pasteurization steps). Our XDO results also do not align with Zhang et al.’s 2022 study, which found XO activity was significantly reduced post-spray drying and freeze drying of raw serum protein concentrate. This differing result could arise from the different sample preparation. LPO, an antimicrobial enzymatic protein (Koksal et al., 2016), was identified in our samples, at approximately 0.1–0.6% of the total abundance (Supplementary Table S7, S8). The relative percentage of LPO (in the TS) was not significantly different between FD-WPC and SD-WPC but was significantly lower in FD-WPC than Feed-WPC (Figure 7e). A follow-up to this study could examine the activity of these enzymes (vs. retention) to determine whether spray drying is more detrimental than freeze drying. As for LA, FABP-H and BTNIA-1, there have been no comparisons in the retention of these proteins (derived from bovine milk) via freeze drying and spray drying in literature. However, Zou et al. (2022) observed a significant decrease of LA in camel milk powder post-spray drying, but not after freeze drying. Likewise, Zhang et al. (2016) observed a significant reduction of FABP-H and LA (100% and 97%, respectively) in their spray-dried bovine milk powder from their raw milk control but did not compare to freeze-dried samples. These findings do not align with the findings from our study, as we found no significant difference between Feed-WPC and SD WPC for these proteins (Figure 7-8).

The peptides with identified Maillard modifications were found to be a minor portion of the total identified peptides, equaling to less than 5% of peptide abundance across all samples examined. Overall, hexose (Hex) and 2-hexose (Hex-2) modifications made up the highest proportion of Maillard modifications to peptides in both the TS and SFS samples. SD-WPC had higher average summed percentage of peptides with Maillard modifications Hex in SFS and Hex (2) in TS and SFS than Feed-WPC (Table 3). Of these, FD-WPC had significantly higher average summed abundances for Hex (2) in SFS than Feed-WPC (Table 3). The average total of all Maillard modifications was significantly higher in both the FD and SD-WPC samples for the SFS group. Overall, SD-WPC had a numerically higher, but not significant, percentage of Maillard modifications than both Feed and FD-WPC (Table 3).

Table 3:

Average summed abundances (%)¹ of the identified Maillard modifications carboxyethyl, carboxymethyl, hexose (Hex), 2-hexose (Hex (2)) and pyrraline on all proteins in the control (Feed-WPC), freeze-dried (FD-WPC) and spray-dried whey protein concentrate (SD-WPC) by total sample (TS) and soluble fraction sample (SFS) (n=3 per sample type).

Modification	Treatment	Feed-WPC	FD-WPC	SD-WPC
Carboxyethyl	TS	0.10 ± 0.06 %a	0.12 ± 0.04 %a	0.12 ± 0.04 %a
	SFS	0.06 ± 0.05 %a	0.08 ± 0.07 %a	0.06 ± 0.07 %a
Carboxymethyl	TS	0.01 ± 0.00 %a	0.09 ± 0.04 %a	0.21 ± 0.19 %a
	SFS	0.07 ± 0.08 %a	0.11 ± 0.05 %a	0.15 ± 0.12 %a
Hex	TS	0.04 ± 0.01 %a	0.90 ± 0.67 %a	1.64 ± 1.34 %a
	SFS	0.04 ± 0.01 %b	0.80 ± 0.11 %ab	1.20 ± 0.55 %a
Hex (2)	TS	0.10 ± 0.03 %b	0.97 ± 0.43 %ab	1.37 ± 0.63%a
	SFS	0.07 ± 0.00 %b	0.83 ± 0.46 %a	0.82 ± 0.07 %a
Pyrraline	TS	0.00 ± 0.00 %a	0.00 ± 0.00 %a	0.01 ± 0.01 %a
	SFS	0.00 ± 0.00 %a	0.01 ± 0.00 %a	0.01 ± 0.001 %a
Total	TS	0.25 ± 0.09 %a	2.08 ± 1.12 %a	3.33 ± 2.00 %a
	SFS	0.24 ± 0.03 %b	1.82 ± 0.62 %a	2.23 ± 0.67 %a

¹Averages with differing superscripts (a,b) do not share significance (Tukey’s HSD, p < 0.05). Values are expressed as the average ± standard deviation.

The observed Maillard reaction products are due to interactions between the residual reducing sugar lactose in the product with the amino acids in the proteins. The final SD-WPC product produced by Hilmar, which typically contains approximately 5.5% residual lactose, may undergo these reactions more rapidly due to high temperatures during spray drying (Table 1; Gómez-Narváez et al., 2019). Schmitz-Shug et al. (2013) found that 10.4% of lysine was lost after spray drying at pilot scale, whereas we identified less than 5% total modifications. However, these values cannot be directly compared as our assessment is based on percent ion abundance, whereas theirs is a quantitative measure specific to lysine. Schmitz-Shug et al. (2013) speculated that some of the observed lysine loss in spray-dried powders could be due to residual moisture content in the final powder during storage, which is a known catalyst for the Maillard reaction for milk protein powders. A limitation of our study is that we did not control for storage temperature of the powders during transportation. Therefore, these changes could be the result of confounding environmental factors (e.g., humidity and temperature fluctuations during storage). Future studies could investigate the differences in Maillard modifications further by measuring lysine, browning index and reducing sugar composition before and after drying. Additionally, these studies could control the conditions in which the powder is stored, minimizing external variables that may accelerate the Maillard reaction.

The methods we used to extract the soluble fraction impacted the abundance, diversity and type of proteins extracted from each sample type, as visualized by Figures 7-8. Some proteins, such as OPN, GLYCAM-1, α-La and BSA, had higher relative percentage in the SFS (Figure 7b) than in the TS (Figure 7a). A possible explanation is that the SFS extraction method reduced the matrix effect from other proteins present in the TS, therefore, allowing the increased detection of heat-stable proteins via increased ionization efficiency.

The two thermal treatment steps (the first applied to milk and the second applied to the incoming whey feed) prior to evaporation likely affected our results (Figure 1). Thermal pasteurization can denature and cause protein aggregation. For example, Wang et al. (2019) reported a significant reduction in MFGM proteins (via LC-MS/MS) in preheated bovine milk (50–70 °C). Additionally, Zhang et al. (2016) saw a significant decrease in the bovine proteins β-Lg, XDO, FABP-H, LA and LF after heat treatment at 62–63 °C for 30 m. The potentially large degradation of whey proteins due to pasteurization steps prior to the drying methods could have limited our ability to detect differences between the two methods. Another possible cause of denaturation could be the ultrafiltration step applied after pasteurization, as the shear and thermal forces involved may impact these proteins. However, previous literature indicates that ultrafiltration and diafiltration do not lead to a decrease in protein composition or cause aggregation in whey (De la Fuente et al., 2002). Further research should investigate the impact of HTST pretreatments on the retention of these bioactive proteins during commercial WPC production.

4. Conclusion

The retention of bioactive whey proteins can be influenced by various processing methods. Although some studies have been conducted on this topic, they have primarily been limited to pilot-scale studies, focusing on specific proteins or alternative mammalian sources. The novelty of this study is that it examined the impact of commercial-scale drying on bioactive whey proteins, offering a more accurate representation than lab or pilot-scale studies. Furthermore, the study is novel as it comprehensively assessed the effects of these processing methods on an array of proteins, rather than focusing on just one. Though SD-WPC had significantly lower average change in enthalpy and average percent solubility than FD-WPC powders, indicating a higher level of denaturation, this change was small and may not be practically meaningful. There was no significant difference in the ANS-based hydrophobicity index between SD- and FD-WPC. Additionally, no significant differences were observed in the levels of some thermo-labile proteins (e.g., LF, IgG and LPO) between SD and FD-WPC powders, as measured by ELISA, SDS-PAGE and LC-MS/MS. We hypothesize that the lack of differences in protein retention between SD- and FD-WPC observed herein is due to the two HTST pasteurization steps that occurred prior to drying, which likely caused initial degradation of many of these proteins. There was a numerically higher summed abundance of Maillard modifications (such as hexose and 2-hexose) in SD-WPC than Feed and FD-WPC. Overall, our results indicate that for commercial dairy processing that involves multiple rounds of pasteurization, freeze drying does not meaningfully improve retention of bioactive proteins compared with spray drying. Further research is warranted to investigate the impact of upstream HTST pasteurization processes prior to spray drying and freeze drying to examine the point at which key bioactive proteins are retained or lost.

Practical Application:

This study aimed to provide insight into the impacts of spray drying vs. freeze drying on whey proteins. Overall, our results indicate that for commercial dairy processing that involves multiple rounds of pasteurization, freeze drying does not meaningfully improve retention of bioactive proteins compared with spray drying. These findings may help the food and dairy industry make informed decisions regarding the processing of its whey protein products to optimize nutritional value.

Nomenclature

ACN

Acetonitrile

ANOVA

Analysis of Variance

ANS

Anilinonaphthalene-1-sulfonic Acid

BCA

Bicinchoninic Acid

BSA

Bovine Serum Albumin

BTN1A1

Butyrophilin subfamily-1 member A1

CEL

Carboxyethylation

DSC

Differential Scanning Calorimetry

DTT

Dithiothreitol

ELISA

Enzyme-Linked Immunosorbent Assay

FA

Formic Acid

FD-WPC

Freeze-dried Whey Protein Concentrate

GLYCAM-1

Glycosylation-Dependent Cell Adhesion Molecule-1

H-FABP

Fatty acid binding protein, heart

Hex

Hexose

Hex (2)

2-Hexose

HPLC

High-Performance Liquid Chromatography

HTST

High-Temperature Short-Time

IgA

Immunoglobulin A

IgG

Immunoglobulin G

IgM

Immunoglobulin M

LA

Lactadherin

LF

Lactoferrin

LC-MS/MS

Liquid Chromatography-Tandem Mass Spectrometry

LPO

Lactoperoxidase

MS

Mass Spectrometry

PAGE

Polyacrylamide Gel Electrophoresis

RFI

Relative fluorescent intensity

RPM

Revolutions Per Minute

SDS-PAGE

Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis

SFS

Soluble Fraction Sample

T_in

Inlet temperature

T_out

Outlet temperature

TFA

Trifluoroacetic Acid

TS

Total Sample

UPLC

Ultra-Performance Liquid Chromatography

WPC

Whey Protein Concentrate

WPI

Whey protein isolate

XDO

Xanthine Dehydrogenase Oxidase

α-La

Alpha-lactalbumin

β-Lg

Beta-lactoglobulin