Efficient in vitro digestion of lipids and proteins in bovine milk fat globule membrane ingredient (MFGMi) and whey-casein infant formula with added MFGMi

The relative immaturity of the infant digestive sys-tem has the potential to affect the bioavailability of dietary lipids, proteins, and their digested products. We performed a lipidomic analysis of a commercial bovine milk fat globule membrane ingredient (MFGMi) and determined the profile of lipids and proteins in the bioaccessible fraction after in vitro digestion of both the ingredient and whey-casein-based infant formula without and with MFGMi. Test materials were digested using a static 2-phase in vitro model, with conditions simulating those in the infant gut. The extent of digestion and the bioaccessibility of various classes of neutral and polar lipids were monitored by measuring a wide targeted lipid profile using direct infusion–mass spectrometry. Digestion of abundant proteins in the ingredient and whey–casein infant formula containing the ingredient was determined by denaturing PAGE with imaging of Coomassie Brilliant Blue stained bands. Cholesterol esters, diacylglycerides, triacylglycerides, phosphatidylcholines, and phosphatidylethanolamines in MFGMi were hydrolyzed readily during in vitro digestion, which resulted in marked increases in the amounts of free fatty acids and lyso-phospholipids in the bioaccessible fraction. In contrast, sphingomyelins, ceramides, and gangliosides were largely resistant to simulated digestion. Proteins in MFGMi and the infant formulas also were hydrolyzed efficiently. The results suggest that neutral lipids, cholesterol esters, phospho-lipids, and proteins in MFGMi are digested efficiently during conditions that simulate the prandial lumen of the stomach and small intestine of infants. Also, supplementation of whey-casein-based infant formula with MFGMi did not appear to alter the profiles of lipids and proteins in the bioaccessible fraction after digestion.


INTRODUCTION
Lipids account for approximately 50% of the energy secreted in human breast milk.The lipids are located in triacylglyceride (TG)-rich milk fat globules (MFG) that are secreted by the lactating mammary epithelial cells.The fat globule is surrounded by a single inner membrane derived from the endoplasmic reticulum, and an outer double membrane originating from the plasma membrane of mammary epithelial cells (Brink and Lonnerdal, 2020).The outer bilayer is rich in phospholipids, cholesterol, sphingolipids, glycophospholipids, and integral membrane proteins.Some peripheral proteins associated noncovalently with the outer membrane of the particle.The intact particle is referred to as the MFG.In addition to its role as an important source of energy for infants, the lipids and proteins in the MFG and metabolites generated during their digestion have diverse health-promoting activities (Brink and Lonnerdal, 2020;Fontecha et al., 2020;Silva et al., 2021).Such activities include the maturation of the gut and its microbiota, host defense, neurological and behavioral development, vision, and cognition.The chemical composition of the MFG is dependent on species of origin, stage of lactation, and method of processing (Pinto et al., 2014;Liang et al., 2017;Brink et al., 2020;Yuan et al., 2020) To mediate their beneficial effects, the majority of complex lipids and proteins in the MFG must be digested.This requires that these compounds be liberated from their matrix; be hydrolyzed to lower molecular weight products that partition in the soluble fraction of chyme (the "bioaccessible" fraction); and be transported to gut epithelial cells, where they may be further metabolized, used, and absorbed, and distributed to peripheral tissues.Some of the ingested compounds and their metabolites may also be metabolized by the gut microbiota to produce various health-promoting compounds such as short chain fatty acids.Digestive conditions in the gastrointestinal lumen of the infant differ qualitatively and quantitatively from those in the mature gut (Bourlieu et al., 2014;Nguyen et al., 2015;Poquet and Wooster, 2016).For example, the preprandial pH is higher, the relative extent of lipid hydrolysis is greater, and proteolytic activity is decreased in the stomach of infants compared with those in the mature gut.Likewise, the activities of pancreatic digestive enzymes and the concentrations of bile salts in the small intestinal lumen of infants ingesting breast milk and formula are less compared with those in adults.Moreover, pancreatic-like lipases 1 and 2 and bile salt stimulated lipase, not pancreatic triacylglyceride lipase, are the predominant lipases in the small intestine of infants (Lindquist and Hernell, 2010).The gut microbiota found in infants and adults differ markedly from that in children and adults (Avershina et al., 2016).
There have been numerous in vivo and in vitro investigations of the digestion of lipids and proteins in human milk, bovine milk, and infant formula in the mature gut (Nguyen et al., 2015;Poquet and Wooster, 2016).The assessment of the digestion of lipids in milk, infant formula, and other foods is generally measured by the kinetics and extent of the release of free fatty acids (FFA), whereas the hydrolysis of proteins is determined using either denaturing PAGE or one of several commonly used chemical methods (Rutherfurd, 2010).Because commercial preparations of the milk fat globule membrane ingredient (MFGMi) are readily available, it is being added to some infant formulas (Hernell et al., 2016;Brink and Lonnerdal, 2020;Fontecha et al., 2020;Silva et al., 2021).Information about the digestion of MFGMi itself and the effect of supplementation of some infant formulas with MFGMi on the digestion of lipids and proteins in formulas is needed.The first aim of our investigation was to examine the digestion of the various lipid classes and proteins present in one such product of MFGMi using a static 2-phase model that simulates the conditions in the stomach and small intestine of infants.The second aim was to determine whether the addition of the MFGMi to a whey-casein-based infant formula affected the in vitro digestion of lipids and proteins.

MATERIALS AND METHODS
Because no human or animal subjects were used, this analysis did not require approval by an Institutional Animal Care and Use Committee or Institutional Review Board.

Materials
Lacprodan MFGM-10 was purchased from Arla Foods.Lipid content represented 17.4 ± 0.6% of the mass of MFGMi as determined by gravimetric analysis after extraction, using a modified Folch procedure (Folch et al., 1957;Pikul et al., 1985).The protein content of the ingredient was 72.6 ± 0.6% as determined by measuring the nitrogen content using the Dumas combustion method (Jimenez and Ladha, 1993) with the Elementar Protein Analyzer (Elementar Analysensysteme GmbH).Cow's milk-based infant formula (IF) without added MFGMi (IF -MFGMi) and cow's milk-based infant formula with 6% (wt:wt) MFGMi added (IF + MFGMi) were prepared according to the manufacturer's instructions (Mead Johnson Nutrition).Rabbit gastric extract (RGE15) was purchased from Lipolytech.The mass ratio of gastric lipase to pepsin was 1.0 to 9.8; specific activities of gastric lipase and pepsin were 16.5 ± 2 U/mg and 534 ± 58 U/mg, respectively.Electrophoretic gels, buffers, Coomassie dye, and molecular weight markers were purchased from Bio-Rad.Other reagents and solvents were obtained from Thermo Fisher Scientific unless otherwise noted.

In Vitro Digestion
Three independent samples of bovine MFGMi and both infant formulas were prepared.In vitro digestions were examined using a static 2-phase model that simulated the conditions in the stomach and small intestine of infants (Bourlieu et al., 2014;Poquet and Wooster, 2016;Nguyen et al., 2016;Ménard et al., 2018).Briefly, 250 mg of MFGMi was solubilized in 5 mL of simulated gastric fluid (Minekus et al., 2014) and the volume was increased to 10 mL (pH, 4.0).Similarly, 2.5 mL of either IF -MFGMi or IF + MFGMi was adjusted to 10 mL with simulated gastric fluid (pH, 4.0).Each reaction tube contained 27 mg of RGE15 (44 U gastric lipase and 1,441 U of pepsin/mL) that was incubated with shaking (95 rpm) at 37°C for 45 min.The pH was adjusted to 6.0 with simulated small intestinal fluid (Minekus et al., 2014)   pancreatin (200 units of protease/mg, 200 units of amylase/mg, and 16 units of lipase units/mg) and bile salts (total concentration, 4.2 mM) in a final reaction volume of 30 mL (pH, 6.5).This combination of RGE and pancreatin is an appropriate mixture for investigating the in vitro lipolysis of a test meal compared with either human gastric and pancreatin juices or purified human gastric and pancreatic lipases (Capolino et al., 2011).Both rabbit and human gastric lipases have the same stereopreference for hydrolyzing the sn-3 position of TG, unlike the microbial acid lipases commonly used for in vitro gastric digestion (Sams et al., 2016).Bile salts present most often in the infant gut (Von Bergmann et al., 1975) were used during the small intestinal phase of digestion rather than bile extract to eliminate introduction of exogenous cholesterol, phospholipids, and proteins.The final concentrations of bile salts were 1.8 mM cholic acid, 1.8 mM chenodeoxycholic acid, 0.2 mM deoxycholic acid, 0.2 mM taurocholic acid, and 0.2 mM ursodeoxycholic acid.Reaction tubes were incubated at 37°C with shaking for 120 min.Upon completion of the small intestinal phase, chyme was centrifuged (106,790 × g, 4°C, 35 min) to separate the aqueous and particulate fractions.The aqueous fraction of the digesta was recovered with a syringe and needle, and filtered immediately (0.22-µm pores).The percentage of compounds in the filtered aqueous fraction from predigested MFGMi and the IF is referred to as being bioaccessible [i.e., available for apical uptake by small intestinal absorptive epithelial cells (Chitchumroonchokchai and Failla, 2018)].

Lipid Analysis
Three independent replicates of 100-µL aliquots of predigested and the aqueous fraction from digested MFGMi, IF -MFGMi, and IF + MFGMi, as well as samples containing gastric and small intestinal digestive enzymes, were spiked with 54 deuterium-labeled internal standards (Sciex and Avanti Polar Lipids).Total lipids were extracted using the modified Folch method (Folch et al., 1957;Pikul et al., 1985).The extracts were concentrated to dryness and resuspended in 250 µL of a mixture of methanol and dichloromethane (50:50) containing 10 mM ammonium acetate.The samples were analyzed using the Lipidyzer platform (Sciex) by flow-injection analysis upon direct infusion (8 µL/min for 10 min per sample across positive and negative ionization) into the Turbo V source of a Sciex 5500 QTRAP mass spectrometer equipped with a SelexION differential ion mobility unit as previously described (Ubhi, 2018;Dirajlal-Fargo et al., 2021).Three blanks and 3 quality control (human plasma) samples were also analyzed in the same batch to moni-tor background and ensure consistency.Analysis of the human plasma control produced 730 distinct signals that included 600 annotated lipids.System suitability tests for sensitivity and reproducibility were passed successfully before sample analysis.The following 13 classes of lipids were profiled and semiquantitated using optimized transitions (MS/MS) for the targeted lipid species in addition to the stable isotope-labeled internal standards (Avanti Polar Lipids): cholesterol esters (CE), ceramides (CER), diacylglycerides (DG), dihydroceramides (DCER), FFA, hexosylceramides (HCER), lactosylceramides (LCER); lysophosphatidylcholines (LPC), lysophosphatidylethanolamines (LPE), phosphatidylcholines (PC), phosphatidylethanolamines (PE), sphingomyelins (SM), and TG.
Gangliosides (GS) were extracted using a modification of the method of Schnabl et al. (2009).Briefly, 2-mL solutions of predigested MFGMi, the 2 IF, and the bioaccessible fractions produced during their digestions were added to 8 mL of MeOH: CHCl 3 (1:2) in a glass vial and vortexed at 1,500 rpm at 4°C for 2 h.Then, 6 mL of MeOH: H 2 O (2:1) was added and mixed well before centrifugation at 3,000 rpm at 4°C for 10 min.Supernatant was transferred to a clean tube.The precipitate was re-extracted with 6 mL of MeOH: H 2 O (2:1), mixed, and centrifuged again.The supernatants were combined, and GS were isolated using a Sep-Pak C18 cartridge that had been conditioned by washing with absolute MeOH followed by 50% MeOH.After loading the extracted sample, the cartridge was washed with 30% MeOH, and GS were eluted with 100% MeOH.Eluate was dried under a stream of nitrogen gas (99%) and analyzed using liquid chromatography (LC)-MS using a modification of the method of Rivas-Serna et al. (2015).Samples were analyzed with an ultrahighperformance LC-MS/MS system (1290 Infinity II-6495 MS, Agilent Technologies).A CSF C18 column (Waters Corporation) was used (2.1 × 150 mm,1.8 µm).Mobile phase A consisted of water: isopropanol (50:50), and mobile phase B was methanol.Mobile phases A and B each contained 5 mM ammonium acetate and 0.05% acetic acid.The flow rate was 0.3 mL/min and temperature was 50°C.The initial condition was 70% B and increased linearly to 100% over 8 min, was held for 3.1 min, and the system was re-equilibrated over 2.4 min for a total runtime of 13.5 min.Eluent was sprayed into the MS via an electrospray probe in negative ion mode with source parameters as follows: 200°C drying gas at an 8-L/min flow with a 32-psi nebulizer, 340°C sheath gas at a 12-L/min flow, 3,000-V capillary, 1,000-V nozzle, 150 very high-pressure radio frequency, and 60-V low-pressure radio frequency.Doubly charged precursor ions were the primary form and MS-MS fragmentation was optimized to sialic acid product ions (m/z 290.09).Although the initial method captured all C12 to C26 species in the GD3 standard (even and odd carbon chains with a d18:1 or d18:0 backbone), only the 10 most abundant GD3 species were detected in test samples, so the analysis was focused on these species to maximize sensitivity (duty cycle).C16, C16 DH, C21, C21 DH, C22, C22 DH, C23, C23 DH, C24, and C24 DH [names correspond to saturated fatty acid chain lengths and DH indicates dihydro form (e.g., d18:0)] were monitored by precursor ions 720.90, 721.91, 755.94, 756.95, 762.95, 763.95, 769.96, 770.97, 776.97, and 777.97, respectively, and the collision energy was 30 eV for the C16 species and 35 eV for the larger GD3 species.A cycle time of 720 ms was used for dynamic multiple reaction monitoring acquisition.GM3 and GD3 GS standards were purchased from Avanti Polar Lipids.
Results of lipid analyses are expressed as the concentration (nmol/g) of lipids grouped by class and converted to nanomoles per gram MFGMi and nanomoles per milliliter for IF without and with (wt:wt) MFGMi.

Protein Analysis: SDS-PAGE
Three replicate samples of MFGMi, the 2 infant formulas, and RGE plus bovine pancreatin before and after simulated digestion were analyzed as previously described (Laemmli, 1970), with some modifications.Test samples were mixed with an equal volume of 2× Laemmli loading buffer containing β-mercaptoethanol (9:1, vol/vol).The mixtures were heated at 60°C in a digital dry bath (Thermo Fisher Scientific) for 5 min and cooled to room temperature before loading 30 µL into the wells of stain-free, precast gels (4% to 20%).The internal and external components of the electrode assembly cell were filled with 1× Tris-SDS buffer.All-Blue reagent (3 µL) was used to monitor molecular weight markers.Voltage (90 V) was applied across the gel for the initial 20 min, followed by 160 V for an additional 40 min.The gels were developed using the Chemidoc-Touch Imaging System (Bio-Rad).ImageLab software (v6.1, Bio-Rad) was used to analyze the developed gels from 3 independent samples each of pre-and postdigested MFGMi, IF -MFGMi, and IF + MFGMi.This facilitated quantification of the proportional relationship between band intensity and sample concentration for each lane.The amount of protein in each band was estimated with reference to a standard curve prepared with different concentrations of BSA analyzed by SDS-PAGE followed by staining with All-Blue dye.The edges of a peak were defined as the intensity of approximately 0.01% of selected area volume (García-Cano et al. 2021).Bands corresponding to the molecular mass of 52 to 54 kDa from stained gels (Figures 2 and 4) that separated proteins in samples of pre-and postdigested enzyme control (RGE plus pancreatin without either MFGMi or IF) were excised for sequencing analysis at the Mass Spectrometry and Proteomics Core Facility at The Ohio State University.The bands were subjected to in-gel trypsin digestion and the resulting peptides were analyzed by capillary LC-MS/MS.Peptide fragments generated by MS/MS were compared with the MASCOT database to obtain the amino acid sequence.Proteins with at least 2 matching peptide fragments reported in the UniProtKB database were considered reliable for identification.

Statistics
Means ± standard deviation for 3 independent samples of predigested MFGMi and the tested IF, as well as the bioaccessible fraction of the digested samples are presented.Means were compared for significant differences of P ≤ 0.05 by paired t-test (SAS v.9.1, SAS Institute Inc.).

RESULTS
Milk fat globule membrane ingredient, IF -MFGMi, and IF + MFGMi were digested in vitro using conditions similar to those in the infant stomach and small intestine (see Materials and Methods).The relative concentrations of neutral and polar lipids in predigested MFGM and the bioaccessible fraction after simulated digestion were compared.Similarly, the profile of abundant proteins before and after in vitro digestion of MFGMi alone and the 2 IF were analyzed by comparing their profiles in denatured polyacrylamide gels after staining.

Digestion of Lipids in MFGMi
The profile of predigested MFGMi obtained with the lipidomic platform registered 760 annotated features that were distributed across 13 lipid classes.These included the following with, the numbers in parentheses indicating total annotated species within the class: CE (n = 23); pooled ceramide classes (CER+, n = 22), DG (n = 40), long-chain free fatty acids (LCFFA, n = 24), LPC (n = 10), LPE (n =9), PC (n = 51), PE ( n = 144), SM (n = 12), and TG (n = 437).The range of chain length of the species surveyed was as follows: C12 to C22 for DG, PC, and TG; C12 to C24 for CE, LPC, and LPE; C14 to C22 for PE; C12 to C24 for FFA and SM; and C14 to C26 for CER, DCER, HCER, and LCER.Because the concentrations of CER, DCER, HCER, and LCER were very low, they have been combined and are presented as CER+ (Figure 1).The analytical platform used in this study did not measure free cholesterol, monoacylglycerides (MG), short-and medium-chain FFA, phosphatidylinositols, phosphatidylserines, and GS.
The relative abundance of lipid classes in predigested MFGMi was TG ≫ PE, LCFFA > PC, SM, DG, CE > CER+, LPC, LPE (Figure 1).In vitro digestion resulted in greater than 97% decreases in the concentrations of TG, DG, CE, and PE, and 86% of PC that partitioned in the bioaccessible fraction.This digestion was associated with marked increases in the quantities of bioaccessible LCFFA, LPC, and LPE upon completion of the small intestinal phase of digestion.Total ceramides, SM, and GS were relatively resistant to simulated digestion, with recoveries of 97%, 89%, and 87% of the predigested quantities, respectively, in the bioaccessible fraction of chyme.
Gangliosides were extracted and analyzed separately from the other lipid classes.Twenty-four distinct GD3 compounds were detected and accounted collectively for greater than 90% of total GS in predigested MFG-Mi.GM3 compounds represented the remainder of this class, whereas GT3 compounds were not detected.To optimize sensitivity, analysis was limited to the 10 most abundant compounds in the GM3 family.The relative concentrations of CER backbones were 2 to 5 times greater than DCER.C18:1 and C18:0 were the most common amide linkages for CER and DCER, respectively.The most abundant fatty acid moieties among GD3 compounds were C16:0 and C22:0, with lesser concentrations of C24:0, C23:0, and C21:0.Mean bioaccessibility of the abundant GS species, such as that of CER+ and SM, was not significantly decreased during simulated digestion of MFGMi (Figure 1).

Digestion of Proteins in the MFGMi
Polypeptides with molecular masses equivalent to several of the more abundant proteins in predigested MFGMi were present in the stained gels (Figure 2, lane 1).These include xanthine oxidase (150 kDa), CD36 (76 to 78 kDa), butyrophilin (67 kDa), and α-LA (14 kDa).Bands with molecular masses of caseins and β-lactoglobulin also were also present in predigested MFGMi, confirming the presence of adherent proteins in the processed commercial ingredient.After in vitro digestion of MFGMi, these bands largely disappeared (Figure 2, lanes 2 and 3).However, an intense band with a molecular mass of 54 kDa remained in the chyme and bioaccessible fraction of digested MFGMi.To determine whether this band represented a protein or hydrolyzed product from digested MFGMi or rather the mixture of digestive enzymes, the profile of proteins in the digestive enzyme mixture without MFGMi was examined before and after gastrointestinal digestion (Figure 2, lanes 4 to 6).The band at 54 kDa, along with other less abundant proteins with molecular masses between 28 and 48 kDa in the gastric and pancreatic enzyme extracts, were also resistant to simulated digestion.Liquid chromatography MS/MS sequencing of peptide fragments from the trypsinized band with a molecular mass of 54 kDa in the bioaccessible fraction of the digestion enzyme mixture (lane 6) identified this band as pancreatic α-amylase.Similarly, sequencing trypsin generated peptides from the 28-kDa band in the bioaccessible fraction of the digested mixture of RGE plus pancreatin without MFGMi (lane 6) identified the protein as pancreatic chymotrypsin-like elastase family member 1.

Digestion of Lipids in IF -MFGMi and IF + MFGMi
Lipidomic analysis showed that IF + MFGMi had higher concentrations of CE, GS, PC, PE, and TG compared with those in IF -MFGMi (Figure 3).The rank order of abundance of lipid classes in the 2 predigested infant formulas was TG ≫ DG > LCFFA, PC, PE, SM > CE > GS > CER+.Simulated digestion decreased the concentrations of CE, DG, PC, PE, and TG extensively, and increased the concentrations of LCFFA, LPC, and LPE in both formulas (Figure 3).Despite several differences in the low concentrations of some LCFFA in the 2 formulas before digestion (e.g., C18:1, C18:2, and C20:5), there were no significant differences in the concentration of each LCFFA in the bioaccessible fraction after completion of the small intestinal phase of digestion (Table 1).However, as observed with the MFGMi (Figure 1), the total content of LCFFA after digestion of both formulas was less than predicted for the lipolysis of TG, DG, CE, and phospholipids (Figure 3).This was a result of a failure to measure the amounts of short-and medium-chain FFA and MG.Infant formula -MFGMi contained very low concentrations of free C20:4 (arachidonic acid) and C22:6 (docosahexaenoic acid)-fatty acids that are crucial for the continued development of the infant Significant differences (P < 0.05) between means for each of the predigested samples and its bioaccessible fraction were determined by paired t-test and are indicated by the presence of different superscript letters.brain (Hahn et al., 2020).However, both IF used in our study were supplemented with esterified C20:4 and C22:6 fatty acids.Although the lipidomic platform did not detect the esterified fatty acids in the predigested formulas, these fatty acids were present in the bioaccessible fraction after digestion.Unlike the results with the MFGMi, an inexplicable partial reduction in the concentration of SM and slight, but not significant (P > 0.05) increases in CER+ in the bioaccessible fractions were noted after digestion of both formulas (Figure 3).The additional complement of GS after digestion resulted from the presence of low concentrations of GS in the rabbit gastric and bovine pancreatin extracts.Together, these results suggest that the addition of the MFGMi to the IF did not affect lipolysis of the neutral and phospholipid classes significantly.

Digestion of Proteins in IF -MFGMi and IF + MFGMi
The electrophoretic profiles of the 2 IF before digestion were very similar, with intense bands present at 34, 18, and 14 kDa, representing casein and the whey proteins β-LG and α-LA, respectively (Figure 4, lanes 2 and 5).The slightly greater intensity of major bands in the IF containing MFGMi was a result of the presence of 8% greater quantity of total protein loaded into the well as assessed by the bicinchoninic acid assay.Less intense bands were also present between molecular masses of 50 and 75 kDa.All these bands were largely absent in chyme (lanes 3 and 6) and the filtered aqueous fractions (lanes 4 and 7) after digestion of the 2 IF.Digested IF contained an intense band with a molecular mass of 54 kDa, along with less intense bands with molecular masses of 28 to 48 kDa and a large, diffuse band containing species with molecular masses less than 10 kDa.The bands with molecular masses of 54 and 28 kDa in the digested formulas appear to be identical to those in digested MFGMi (Figure 2, lanes 4 to 6); in other words, pancreatic α-amylase and chymotrypsinlike elastase family member 1.Because there were no clearly identifiable bands of MFGMi proteins in the complex matrix of the 2 IF because of their relatively low amount in the supplemented formula, it is unknown whether MFGMi proteins in IF + MFGMi were hydro- lyzed with the same efficiency as when MFGMi alone was digested (Figure 2).

DISCUSSION
Digestibility studies of IF have focused mainly on the protein fraction for compliance with regulatory agencies and to mimic closely the postprandial amino acid balance of human milk.However, limited information is available on both the digestibility of the lipid fraction of IF and the digestibility of the lipids and proteins in the MFGMi that is being added to some IF.Given the importance of the reported benefits of MFGMi on the development and health of infants, we addressed these issues.
Digestion of dietary lipids is initiated in the infant stomach by gastric lipase.The enzyme hydrolyzes preferentially a portion of the esterified fatty acids at the sn-3 position of the glycerol backbone.Within the small intestine of infants, pancreatic lipase, bile salt pancreatic lipase, and lipase-like protein 2 are largely responsible for hydrolyzing TG at the sn-1 and sn-3 positions, with the latter 2 enzymes partially degrading MG to glycerol and FFA.Long-chain FFA and products of other insoluble dietary lipids partition into bile salt micelles and are delivered to the brush border surface of absorptive epithelial cells, where they are transported into the cytoplasm.Released short-and medium-chain FFA do not require incorporation into micelles for facilitated transport across the brush border membrane of these cells (Poquet and Wooster, 2016).The majority of previous in vitro studies of digestion of IF have focused primarily on the release of FFA as being indicative of the extent of lipolysis.The kinetics and extent of in vitro lipolysis in milk and IF are influenced by their chemical composition, type of processing, and the size of the lipid globules (Liang et al., 2017;Yuan et al., 2020).Our results show that TG, DG, and CE in MFGMi and in the whey-casein-based IF were digested extensively using conditions simulating those in the gastric and small intestinal compartments of the infant (Table 1, Figures 1 and 3).However, the pool of LCFFA generated was less than expected by the extensive digestion of the abundant neutral lipids.The likely explanation is that the lipidomic platform used in our investigation did not analyze short-and medium-chain FFA and MG.This suggestion is supported by reports from several previous investigations.Kopf-Bolanz et al. (2012) reported that TG in pasteurized and homogenized milk were digested completely to FFA and MG during in vitro digestion conditions simulating those in the ma-ture gastrointestinal tract.Hageman et al. (2019) found considerable amounts of short-and medium-chain FFA (C4:0 to C11:0) in digested bovine milk-based infant formula using a static 2 stage in vitro model with adultlike conditions.Lipidomic analysis showed the presence of DG, MG, and FFA in chyme after in vitro digestion of bovine milk fat using adult-like conditions (Teng et al. 2019). Moreover, Pérez-Gálvez et al. (2021) recently reported similar concentrations of MG, DG, and FFA in the bioaccessible fraction after in vitro digestion of pasteurized human milk using conditions simulating those in the infant gastrointestinal tract.
The main classes of polar lipids in milk include glycerophospholipids, sphingolipids, and GS.These lipids have essential roles in cellular membrane structure and function-signal transduction processes associated with cellular growth, differentiation, and apoptosisand the maturation of the infant gut, immune system, and brain (Nilsson and Duan 2019;Ortega-Anaya and Jimenez-Flores, 2019;Jiang et al., 2022).It was also shown recently that milk fat polar lipids contribute to the efficient digestion of the milk fat droplets (Luo et al., 2019).Phosphatidylcholines and PE are the most abundant glycerophospholipids in milk and are degraded to their lysometabolites in the small intestine by pancreatic phospholipases.As expected, the majority of PC and PE in MFGMi and the IF was degraded during in vitro gastrointestinal digestion (Figures 1 and 3).Sphingomyelins are important structural components of cell membranes and participate in some signal transduction processes required for cellular growth, differentiation, and apoptosis.Although human milk is rich in SM, the concentration of SM, like glycerophospholipids, present in standard IF is low.Previous investigation of the digestion of SM in animals and humans showed that apical uptake by small intestinal epithelial cells is limited (Schmelz et al., 1994;Nilsson and Duan, 2006).Dietary SM are hydrolyzed to CE by alkaline sphingomyelinase released from the small intestinal brush border membrane in the mid-and lower regions of the small intestinal lumen.The generated CE are metabolized further to sphingosine and fatty acids by neutral ceramidase-metabolites that are transported into absorptive epithelial cells (Nilsson andDuan 2006, 2019).It was recently reported that SM and its metabolites may affect functionality in the gastrointestinal tract directly.Milk SM increased the expression of mRNA that are translated into tight junction proteins in monolayers of differentiated Caco-2/TC7 human intestinal cell cultures and the expression of tight junction proteins in mouse ileum, possibly via regulation of the expression of IL-8 (Milard et al., 2019).Sphingomyelins were relatively stable during in vitro digestion of both MFGMi and IF enriched with the same MFGMi in our study, likely because the in vitro model was not coupled with an intestinal cell system expressing sphingomyelinase and neutral ceramidase activities (e.g., differentiated Caco-2 human intestinal cells) or within the small intestinal lumen.
Gangliosides are sialic acid-containing glycosphingolipids that reside in cell membranes and are essential for neuronal maturation, nerve transmission, gut barrier function, and the maturation of the gut immune system of the infant (Rueda, 2007;Palmano et al., 2015).As reported previously for human milk and bovine MFGM, GD3 represented the majority of GS in the MFGMi used in our study (Lacomba et al., 2011;Brink et al., 2020).Our results suggest that GD3 compounds were relatively stable during simulated gastric and small intestinal digestion.This finding contrasts with the report of Lacomba et al. (2011), who also used a static 2-phase simulated model of infant digestion of human milk and casein-whey-based IF.The reported losses of sialic acid during gastric digestion of powdered IF and human milk were 23% and 13%, respectively, followed by substantial declines in the content of GD3 during the small intestinal phase of digestion.Further investigation is required to address more critically the fate of dietary GS during digestion.
Previous investigations have shown that results from studies using static and dynamic models of in vitro digestion of proteins in bovine and human milk and whey-casein-based IF are relatively well correlated with in vivo findings in the gastrointestinal tract of adult humans and pigs (Kopf-Bolanz et al., 2012;Ménard et al., 2014;Egger et al., 2016Egger et al., , 2019;;Sanchón et al., 2018).It is recognized that the kinetics and the relative extent of digestion of specific proteins in milk and MFGMi, as well as the gastrointestinal compartment in which their digestion primarily occurs, differ.For example, caseins and α-LA were degraded rapidly during in vitro gastric digestion, whereas proteolysis of β-LG (an endogenous protein in the MFG) occurs during the small intestinal phase of in vitro digestion and in vivo (Kopf-Bolanz et al., 2012;Ménard et al., 2018;Sanchón et al., 2018).Also, butyrophilin, another endogenous protein in the MFGM, was more resistant to peptic digestion than several of the other endogenous proteins in MFGMi, such as xanthine oxidase, PAS 6, and PAS 7 (Ye et al., 2011;Le et al., 2012).In our study, there was extensive degradation of proteins in MFGMi and IF (Figures 2  and 4).As observed by several other investigators, there was a 54-kDa protein present in the chyme, the bioaccessible fraction of digested MFGMi, and the IF that was not present in the predigested materials (Figure 2).This band has been reported by others as a component of pancreatin (Kopf-Bolanz et al., 2012;Ménard et al., 2014Ménard et al., , 2018;;Sanchón et al., 2018;Egger et al., 2019).We have confirmed the finding of Egger et al. (2016) that the protein that resisted simulated digestion is pancreatic α-amylase.In addition, we identified a 28-kDa protein resistant to simulated digestion as pancreatic chymotrypsin-like elastase family member 1.

CONCLUSIONS
Abundant proteins and neutral and most polar lipids in bovine MFGMi and IF supplemented with MFGMi were efficiently digested using in vitro conditions that simulate the infant stomach and small intestine.The relative stability of SM and CER+ during simulated digestion in the absence of intestinal absorptive cells aligns well with in vivo findings.Results suggest that the addition of MFGMi to IF did not impair the digestion of abundant lipids and proteins.Limitations of the present study include the use of the in vitro digestion model does not recapitulate the dynamic nature of in vivo digestion (Nugyen et al., 2016;Bohn et al., 2018), the semiquantitative nature of the lipidomic analysis, and the absence of information on several lipid classes.Future studies should couple addition of the bioaccessible fraction of digested MFGMi to appropriate intestinal cell and organoid models to assess metabolite transport and effects on cell functions.
Chitchumroonchokchai et al.: DIGESTION OF MILK FAT GLOBULE MEMBRANE INGREDIENT Figure 1.Profile of lipid classes in milk fat globule membrane ingredient (MFGMi) before and in the bioaccessible fraction after in vitro digestion using conditions representative of those in the infant stomach and small intestine.CE = cholesterol esters; CER+ = combined ceramides + dihydroceramides + hexosylceramides + lactosylceramides; DG = diacylglycerides; LCFFA = long-chain free fatty acids; GS = gangliosides; LPC = lysophosphatidylcholines; LPE = lysophosphatidylethanolamines; PC = phosphatidylcholines; PE = phosphatidylethanolamines; SM = sphingomyelins; TG = triacylglycerides.Data are mean ± SD for 3 independent experiments.Different letters above bars indicate that the quantities in the pre-and postdigested samples for the listed class of lipids differ significantly (P < 0.05).

1
Data are mean ± SD for 3 independent samples of predigested MFGMi, IF -MFGMi, IF + MFGMi, and their respective bioaccessible fractions after in vitro digestion.Results for MFGMi were converted to concentrations per gram of powdered MFGMi.Results for IF -MFGMi and IF + MFGMi were converted to concentrations per milliliter. 2 ND = not detected.

Figure 2 .
Figure 2. Proteins in milk fat globule membrane ingredient (MFGMi) are efficiently degraded during simulated gastric and small intestinal digestion using prandial-like conditions in the infant gut.Representative SDS-PAGE (4% to 20% cross-linked gels) profiles in predigested MFGMi, chyme, and filtered aqueous (i.e., bioaccessible) fraction after in vitro digestion are shown in lanes 1, 2, and 3, respectively.Labels for abundant proteins in MFGMi are listed to the left of lane 1.The profiles of digestive enzymes (rabbit gastric extract and bovine pancreatin) in the absence of MFGMi before and after simulated digestion are shown in lanes 4, 5, and 6.Lane 7 shows molecular masses (MM) for the protein standard.

Figure 4 .
Figure 4. Proteins in the infant formulas without and with 6% milk fat globule membrane ingredient (MFGMi) are hydrolyzed efficiently during in vitro digestion using conditions that simulate those in the upper gut of infants, and are independent of the presence or absence of MFGMi.Representative SDS-PAGE profiles of proteins in predigested formula, chyme, and filtered aqueous (i.e., bioaccessible) fraction after simulated digestion of IF -MFGMi is shown in lanes 2, 3, and 4, respectively; and of IF + MFGMi in lanes 5, 6, and 7, respectively.The intense band at 54 kDa in lanes 2, 3, 5, and 6 is the same as that of the digestive enzyme control in the absence of infant formula (see Figure2, lanes 4 to 6).Lane 1 contains standard protein markers of known molecular masses (MM).
Chitchumroonchokchai et al.: DIGESTION OF MILK FAT GLOBULE MEMBRANE INGREDIENT

Table 1 .
Chitchumroonchokchai et al.: DIGESTION OF MILK FAT GLOBULE MEMBRANE INGREDIENT Concentration of long-chain free fatty acids (LCFFA) in predigested milk fat globule membrane and in the bioaccessible fraction of digested milk fat globule membrane ingredient (MFGMi), control infant formula (IF -MFGMi), and infant formula containing MFGMi (IF +