**Milk Containing A2** β**-Casein ONLY, as a Single Meal, Causes Fewer Symptoms of Lactose Intolerance than Milk Containing A1 and A2** β**-Caseins in Subjects with Lactose Maldigestion and Intolerance: A Randomized, Double-Blind, Crossover Trial**

**Monica Ramakrishnan, Tracy K. Eaton, Omer M. Sermet and Dennis A. Savaiano \***

Department of Nutrition Science, College of Health and Human Sciences, Purdue University, West Lafayette, IN 47907, USA; ramakrm@purdue.edu (M.R.); tkeaton@purdue.edu (T.K.E.); omer\_sermet@hotmail.com (O.M.S.)

**\*** Correspondence: savaiano@purdue.edu

Received: 12 November 2020; Accepted: 15 December 2020; Published: 17 December 2020

**Abstract:** Acute-feeding and multiple-day studies have demonstrated that milk containing A2 β-casein only causes fewer symptoms of lactose intolerance (LI) than milk containing both A1 and A2 β-caseins. We conducted a single-meal study to evaluate the gastrointestinal (GI) tolerance of milk containing different concentrations of A1 and A2 β-casein proteins. This was a randomized, double-blind, crossover trial in 25 LI subjects with maldigestion and an additional eight lactose maldigesters who did not meet the QLCSS criteria. Subjects received each of four types of milk (milk containing A2 β-casein protein only, Jersey milk, conventional milk, and lactose-free milk) after overnight fasting. Symptoms of GI intolerance and breath hydrogen concentrations were analyzed for 6 h after ingestion of each type of milk. In an analysis of the 25 LI subjects, total symptom score for abdominal pain was lower following consumption of milk containing A2 β-casein only, compared with conventional milk (*p* = 0.004). Post hoc analysis with lactose maldigesters revealed statistically significantly improved symptom scores (*p* = 0.04) and lower hydrogen production (*p* = 0.04) following consumption of milk containing A2 β-casein only compared with conventional milk. Consumption of milk containing A2 β-casein only is associated with fewer GI symptoms than consumption of conventional milk in lactose maldigesters.

**Keywords:** A1 beta-casein; A2 beta-casein; beta-casomorphin; gastrointestinal intolerance; hydrogen breath test; lactose challenge; lactose intolerance symptoms; milk intolerance; Qualifying Lactose Challenge Symptom Score

#### **1. Introduction**

Approximately 30% of cows' milk protein is β-casein [1], of which two genetic variants exist: A1 and A2 [2]. A1 β-casein includes histidine at the 67th position in the peptide chain, whereas A2 β-casein includes proline at this position [3]. Although some cattle breeds maintain the A2 β-casein variant, a single nucleotide polymorphism in modern western cattle breeds means that they exhibit mixed A1 and A2 β-casein variants [3–5]. Digestive enzymes act on A1 β-casein and hydrolyze it, releasing beta-casomorphin-7 (BCM-7) [6–10]. The histidine residue in A1 β-casein allows cleavage to form BCM-7, whereas the proline residue in A2 β-casein limits such cleavage and BCM-7 formation [11].

In animal studies, BCM-7 is both pro-inflammatory and associated with slower gastrointestinal (GI) transit [12]. In intestinal and neuronal cells, BCM-7 downregulates the glutathione (GSH) levels [13], which is an important antioxidant in the body for combating oxidative stress, which otherwise can

result in inflammation [14]. Oxidative stress has been shown to induce epigenetic changes, especially on genes that are important mediators of inflammation, leading to increased GI symptoms.

Multiple-day and acute-feeding studies in Chinese and Australian populations have shown that milk containing only the A2 β-casein protein caused fewer symptoms of lactose intolerance (LI) than milk containing both A1 and A2 β-casein protein [15–18]. However, these studies were conducted in subjects with self-reported LI, and no relevant blinded studies have been reported in verified LI individuals; this is a notable omission, given that self-reported LI can be unreliable [19]. Moreover, only one study to date has examined the effects of the administration of variable ratios of A1 and A2 β-casein [20]. Manifestations of LI can be both acute and chronic: they can be long-term due to epigenetic changes or genetic mutation [21]; they also appear to be a single-meal event, with symptoms occurring between 30 min and 6 h after exposure [22,23]. These symptoms and their cause are distinct from milk allergy, which results from an immune reaction to milk proteins.

We conducted a randomized, double-blind, single-meal feeding trial with four types of milk, which varied in A1/A2 β-casein protein ratio: milk containing A2 β-casein only, Jersey milk (containing 25%/75% A1/A2 β-casein), conventional milk (containing 75%/25% A1/A2 β-casein), and lactose-free milk. The principal study objective was to determine, via a hydrogen breath test (HBT) in lactose maldigester individuals living in the Midwest United States and who had verified LI following a blinded milk challenge, if lactose digestion and GI tolerance were affected by the four different milk types within 6 h after ingestion. We hypothesized that a single meal of milk containing A2 β-casein only would be better tolerated, producing fewer GI symptoms and less maldigestion during the 6 h study, than conventional milk containing both A1 and A2 β-casein. We also hypothesized that Jersey milk would produce an intermediate response regarding the HBT and occurrence of GI symptoms as a result of its higher level of A2 β-casein and lower level of A1 β-casein.

#### **2. Materials and Methods**

#### *2.1. Subject Selection and Inclusion Criteria*

This was a randomized, double-blind, crossover trial in subjects aged 18–65 years. Subjects were recruited through flyers and advertisements in local (West Lafayette, IN, USA) and campus (Purdue University, West Lafayette, IN, USA) newspapers, and Purdue Today email. Study recruitment started in February 2018 and was suspended in February 2020 due to COVID-19 restrictions by our institutional review board (IRB). A total of 853 people indicated interest in the study and contacted the investigators via email or phone. Of these, 258 subjects successfully completed phone screening and were then categorized as eligible or ineligible to participate in the trial (Appendix A). After the phone screens, all eligible subjects signed a consent form and agreed to participate in the study. We queried all interested subjects regarding demographic information, current medication use, and height and weight for calculating body mass index (BMI). Participants were assigned an identification number upon signing the informed consent form. Identification numbers were block pre-randomized using randomization.com. Staff performed the randomization, enrolled participants, and provided milk type information to our Clinical Research Center (CRC) kitchen via sealed security envelopes. Participants and study staff assessing outcomes over the 6 h were blinded to the milk types consumed.

Eligible subjects had to have avoided dairy for at least 1 month prior to screening and were included in the study only if they agreed to refrain from dairy and all treatments or products used for dairy intolerance (e.g., Lactaid® dietary supplements; McNeil Nutritionals, LLC, Ft. Washington, PA, USA) throughout the trial. Subjects had to have a history of perceived dairy intolerance. Perceived LI was then confirmed by the Qualifying Lactose Challenge Symptom Score (QLCSS) duringa6h HBT after consumption of a commercial milk containing a high amount of A1 β-casein [24]. Lactose maldigesters, defined as producing more than 20 parts per million (ppm) hydrogen at any time point following the baseline commercial milk challenge [25], were eligible for the intervention portion of the study.

Abdominal pain, bloating, flatulence, and diarrhea are typical symptoms of LI [26]. Although other studies of A1 and A2 β-caseins reported scores for stool frequency and stool consistency as indicators of LI [15–17], in the current study, qualifying scores for abdominal pain, bloating, flatulence, diarrhea, and fecal urgency were each recorded by subjects using a six-point Likert scale. Symptoms were ranked from 0 to 5, where 0 was for no symptoms, 1 for slight, 2 for mild, 3 for moderate, 4 for moderately severe, and 5 for severe symptoms. If subjects met one of the following three criteria regarding the QLCSS, they were classified as lactose intolerant: a score of 4 or 5 for an individual symptom; a score of 3 for at least two symptoms; or a score of 3 for one symptom at two time points in the study.

#### *2.2. Exclusion Criteria*

Subjects were excluded from the study for the following reasons: allergy to milk; pregnancy or lactation; cigarette smoking, or use of tobacco or nicotine-containing products within 3 months of screening; diagnosis of abnormal GI motility; a history of GI tract surgery; the presence of any medical condition with symptoms that could confound collection of data about adverse events; ulcer; diabetes mellitus; congestive heart failure; HIV, hepatitis B, or hepatitis C virus infection; BMI > 35 kg/m2; use of products to treat dairy intolerance within 7 days of screening; use of antacids and/or proton pump inhibitors; use of antibiotics or colonic enemas within 30 days prior to screening; any concurrent disease or symptoms that may interfere with assessment of the cardinal symptoms of dairy intolerance; use of ethanol (alcohol) and/or drug abuse in the past month; chemotherapy; or use of any investigational drug, or participation in any investigational study, within 30 days prior to screening.

#### *2.3. Interventions*

Four types of milk were evaluated in the study: milk containing A2 β-casein only (The a2 Milk Company, Boulder, CO, USA); Jersey milk (containing 25%/75% A1/A2 β-casein; American Jersey Cattle Association, Crockett, VA, USA); conventional milk (containing 75%/25% A1/A2 β-casein; Kroger® 2% reduced fat; The Kroger Co., Indianapolis, IN, USA); and lactose-free milk (containing 60%/40% A1/A2 β-casein; Lactaid®; McNeil Nutritionals, LLC). Jersey milk was shipped to our study laboratory (Purdue University, West Lafayette, IN, USA) from the American Jersey Cattle Association (Crockett, VA, USA). All other milk products were purchased from Payless (West Lafayette, IN, USA); if they were unavailable at Payless, then they were purchased from Fresh Thyme (West Lafayette, IN, USA). All four milk types were procured every two weeks and administered prior to their expiry dates.

#### *2.4. A1*/*A2 Analysis*

The ratio of A1/A2 β-casein in the four types of milk was analyzed using mass spectrometry (MS) at Purdue Proteomics Facility. Protein extraction from 100 μL of each milk was performed by denaturation with 400 μL of 8 M urea and 10 mM dithiothreitol (DTT) and vortexing for 15 min at room temperature to remove fat. This was followed by protein alkylation with 400 μL of 8 M urea and 10 mM DTT, and then digestion with pepsin using a 1:20 enzyme to substrate ratio for 1 h at room temperature. The peptides, which were cleaned/desalted via a C18 Silica MicroSpin column (The Nest Group Inc., Southborough, MA, USA) after digestion, were analyzed with the Dionex UltiMate 3000 RSLC nano System combined with the Q-Exactive High-Field Hybrid Quadrupole Orbitrap MS (Thermo Fisher Scientific, Waltham, MA, USA). Peptides were then re-suspended in 3% acetonitrile/0.1% formic acid/96.9% Milli-Q water, and 5 μL (1 μg) were used for liquid chromatography (LC)-MS/MS analysis. A trap (300 μm internal diameter (ID) × 5 mm packed with 5 μm 100 Å PepMap C18 medium; Thermo Fisher Scientific) was used to separate peptides and a 120 min gradient method with a flow rate of 300 nL/min was used for the analytical columns (75 μm ID × 15 cm long packed with 3 μm of 100 Å PepMap C18 medium). Mobile phase A contained 0.1% formic acid in water and mobile phase B contained 0.1% formic acid in 80% acetonitrile. The linear gradient started at 5% B and reached 30% B in 80 min, 45% B in 91 min, and 100% B in 93 min. The column was held at 100% B

for 5 min and then brought back to 5% B. The column was held at 5% B for 20 min to equilibrate at 37 ◦C. The top 20 data-dependent MS/MS scan method was used to acquire MS data with a maximum injection time of 100 ms and a resolution of 120,000 at 200 *m*/*z*. High-energy C-trap dissociation with the normalized collision energy of 27 eV was used to fragment precursor ions. MS/MS scans were acquired at a resolution of 15,000 at 200 *m*/*z*. Repeated scanning of identical peptides was avoided by setting the dynamic exclusion at 20 s.

LC-MS/MS data were analyzed using MaxQuant software (version 1.6.0.1; Max Planck Institute of Biochemistry, Martinsried, Germany). The combined non-redundant *Bos taurus* protein sequence database downloaded from UniProt (www.uniprot.org) in January 2017 was used for protein identification and label-free relative quantitation. The following parameters were used for database searches: precursor mass tolerance of 10 ppm; enzyme pepsin allowing up to two missed cleavages; oxidation of methionine as a variable modification and iodoethanol as a fixed modification. The false discovery rate of peptide spectral match and protein identification was set to 0.01. Only proteins with a label-free quantitation value of 0 and MS/MS spectral counts of ≥2 were considered as a true identification before being used for further analysis.

#### *2.5. Sugar, Protein, and Fat Analyses*

Total sugars, fat, and protein were analyzed by Eurofins Food Integrity and Innovation (Eurofins Food Chemistry Testing US, Inc., Madison, WI, USA) [27–29]. The sugar profile was determined using 10 g of each milk type (with the exception of lactose-free milk), and sugars were extracted with a 50:50 methanol:water solution. Inert gas was used to dry each sample, which was derivatized prior to analysis, and the analysis was conducted via gas chromatography with flame ionization detection.

Because of its very low lactose content, lactose-free milk was analyzed using a different procedure. A 10 g milk sample was extracted with dilute HCl and centrifuged. The supernatant was filtered using a strong cation exchange cartridge, and an OnGuard II syringe filter (Thermo Fisher Scientific) was used for neutralization. Applicable amounts of dilutions were injected into a high-performance anion exchange chromatography system equipped with pulsed amperometric detection (Thermo Fisher Scientific).

Fat in the samples was analyzed by base hydrolysis, and protein was analyzed using the Dumas method [27].

#### *2.6. Study Procedures*

The subjects reported to our clinical research facility (Purdue University) for four visits, with at least six days between any two consecutive visits. Subjects consumed a low-fiber dinner and fasted for 12 h prior to visits. Subjects consumed a different randomized milk product at 8 a.m. on the day of each visit. Each milk meal, except the lactose-free milk, contained ~4.5 g of lactose/per 100 mL. The amount of milk consumed was calculated as approximately 0.5 g of lactose times bodyweight in kg, divided by 11 g (the normal amount of lactose in a cup of regular milk), then multiplied by 245 mL (one cup).

$$\text{Amount of milk consumed} = \frac{0.5 \text{ g} \text{ of latose} \times \text{bodyweight (kg)}}{11 \text{ g} \text{ of latose}} \times 245 \text{ mL of milk}$$

Subjects always consumed the same quantity of fluid (mL) and lactose (g), despite small variations in the lactose content.

#### *2.7. Study Endpoints*

The primary study endpoints were the occurrence of GI symptoms (abdominal pain, bloating, flatulence, diarrhea, and fecal urgency), and measurement of hydrogen in breath samples (a standard measure of maldigestion), for up to 6 h after consumption of each milk type. Breath samples were collected, and GI symptoms were recorded by the subjects, at 0, 0.5, 1, 1.5, 2, 3, 4, 5, and 6 h after

commercial milk ingestion. To measure hydrogen in the breath samples, a hydrogen microanalyzer (QuinTron BreathTracker Digital Microlyzer, model SC; QuinTron Instrument Company, Inc., Milwaukee, WI, USA) was used. An increase of 20 ppm hydrogen between any two timepoints in the study indicated lactose maldigestion [26]. Symptoms were scored using a six-point Likert scale (as described in the inclusion criteria).

#### *2.8. Study Ethics*

The study (ClinicalTrials.gov #NCT03713346) and its protocol were approved by the Purdue IRB (IRB #1710019781). The trial was conducted in accordance with the Helsinki Declaration of 1975 as revised in 1983. The study was also conducted in accordance with International Conference on Harmonization Good Clinical Practice guidelines.

#### *2.9. Statistical Analyses*

The initial power calculation indicated a sample size requirement of 26, which was determined based on the selection of a 20% decrease in area under the curve for change in breath hydrogen as the minimal difference that would be clinically significant. Power calculations indicated that completion of the protocol, with a crossover study design, by 26 subjects would be adequate to demonstrate 80% statistical power, consistent with biological relevance using α = 0.05 to detect a 20% change in breath hydrogen. The sample size for symptoms was derived from a previous study in Chinese preschoolers aged 5 to 6 years [18]. The COVID-19 pandemic caused us to suspend the study in February 2020 with 25 verified LI subjects and an additional eight maldigesters who did not meet the LI criteria. However, we observed that not all LI subjects met the criteria for LI after a second commercial milk dose, suggesting that the symptom criteria were arbitrary and inconsistent. A post hoc analysis indicated only 15 of 25 LI subjects met the LI criteria after receiving the second commercial milk challenge as part of the randomized intervention; therefore, we included maldigesters who did not meet the criteria for LI in the study in June 2019, to better understand the potential effect of milk containing A2 β-casein only and Jersey milk on all maldigesters. As a result, an analysis was conducted with 25 LI subjects, and a post hoc analysis was conducted with the addition of eight maldigesters (a total of 33 subjects).

GI symptoms, hydrogen at each timepoint, and total hydrogen during each 6 h study period were analyzed using the paired *t*-test. Two-tailed *p*-values were compared with a significance level of 0.05. Descriptive statistics were used to calculate mean and standard error values for GI symptoms and breath hydrogen. All statistical analyses were conducted using Microsoft® Excel and the Statistical Package for the Social Sciences (IBM SPSS Statistics for Windows, Version 26.0; IBM Corp., Armonk, NY, USA).

Lactose-free milk was used as a negative control. Conventional milk was compared with Jersey milk and milk containing A2 β-casein only, by measuring the occurrence of GI symptoms and breath hydrogen. Subject symptom scores for abdominal pain, bloating, flatulence, and diarrhea were summed over 6 h after each milk consumption, and total symptoms for each subject were calculated as the sum of all the total symptom scores for abdominal pain, bloating, flatulence, and diarrhea; symptoms of fecal urgency were analyzed separately. For each subject, baseline breath hydrogen concentration was subtracted from the breath hydrogen concentration produced at each timepoint (0, 0.5, 1, 1.5, 2, 3, 4, 5, and 6 h) to correct for residual hydrogen.

#### **3. Results**

#### *3.1. Baseline and Demographic Characteristics*

Of the 258 subjects who were phone-screened, 111 were ineligible due to medical conditions or lack of milk avoidance, and five subjects were ineligible because of Lactaid® use. A total of 142 subjects were eligible for the HBT screening, but only 94 chose to participate; the other 48 subjects did not respond to attempts to schedule the baseline screening. A total of 35 subjects met the maldigestion criteria and were randomized to one of four sequences for receiving the four milk products. Thirty-three completed the four study visits and had GI symptoms and hydrogen production in HBTs recorded for 6 h after each milk treatment. Of the 33 subjects, 25 met the symptom criteria and were classified as LI and an additional eight subjects were maldigesters without LI; two subjects were unable to complete the protocol owing to COVID-19 restrictions implemented by our IRB (Figure 1).

**Figure 1.** Study enrollment, randomization and analyses. HBT, hydrogen breath test; LI, lactose intolerance/intolerant.

Overall, 15 male and 18 female subjects with a mean age of 25 (range 19–50) years, and a mean BMI of 24 (range 18–33) kg/m2, completed the study. The study population comprised 14 individuals who identified as Asian, four African Americans, 14 Caucasians, and one American Indian. All participants resided in the United States: five were Hispanic, 26 were non-Hispanic, and two participants did not disclose ethnicity (Hispanic/non-Hispanic) (Table 1). Nutrient composition for each of the four milk products evaluated is shown in Table 2.


#### **Table 1.** Baseline and demographic characteristics.

Baseline and demographic characteristics of maldigesters and lactose intolerant subjects (*n* = 33) enrolled in this randomized, double-blinded trial comparing conventional milk with Jersey milk and milk containing A2 β-casein only. BMI, body mass index; ppm, parts per million; QLCSS, Qualifying Lactose Challenge Symptom Score.



Subjects were fed approximately 4.5 g lactose/100 mL of each milk after an overnight fast, in random order, with six days between treatments.

#### *3.2. GI Symptoms*

#### 3.2.1. LI Subjects

The total symptom score for abdominal pain during the 6 h after the consumption of milk containing A2 β-casein only, in LI subjects (*n* = 25), was significantly lower than that following consumption of conventional milk (112 vs. 146; *p* = 0.004); however, in contrast, the total score for abdominal pain after consumption of Jersey milk was not significantly different from that for conventional milk (135 vs. 146; *p* = 0.63). Regarding total symptom scores for bloating, flatulence, diarrhea, and fecal urgency, no significant differences were evident between those who had consumed conventional milk and those who had consumed either Jersey milk or milk containing A2 β-casein only (Figure 2). With respect to the combined total symptom scores for abdominal pain, bloating, flatulence, and diarrhea reported by subjects, there were no significant differences between conventional milk and Jersey milk or milk containing A2 β-casein only (Table 3 and Figure 2).

#### 3.2.2. All Maldigesters

The total symptom score for abdominal pain after the consumption of milk containing A2 β-casein only, in all maldigesters (*n* = 33), was significantly lower than that following consumption of conventional milk (126 vs. 175; *p* = 0.001); however, the total score for abdominal pain after the consumption of Jersey milk was not significantly different from that for conventional milk (170 vs. 175; *p* = 0.83). Total symptom score for bloating was higher when consuming Jersey milk compared with conventional milk (293 vs. 240; *p* = 0.05). Total symptom scores for flatulence, diarrhea, and fecal urgency were similar in subjects consuming milk containing A2 β-casein only, Jersey milk, and conventional milk (Figure 3). The combined total symptom scores for abdominal pain, bloating, flatulence, and diarrhea showed there were fewer symptoms with milk containing A2 β-casein only (601 vs. 737; *p* = 0.04) compared with conventional milk, whereas consumption of Jersey milk or conventional milk produced similar symptom scores (790 vs. 737; *p* = 0.44) (Table 3 and Figure 3).

**Figure 2.** Total symptoms reported during the 6 h after consuming the four milk products in 25 lactose intolerant subjects. \*\* *p* = 0.004 for abdominal pain due to milk containing A2 β-casein only vs. conventional milk.


**Table 3.** Comparison of total hydrogen produced, and symptoms reported.

Comparison of total hydrogen produced, and symptoms reported for six hours following consumption of conventional milk versus milk containing A2 β-casein only and conventional milk versus Jersey milk using paired *t*-tests. LI, lactose intolerant; ppm, parts per million; <sup>a</sup> abdominal pain + bloating + flatulence + diarrhea.

#### *3.3. HBT Results*

#### 3.3.1. LI Subjects

Hydrogen breath concentration was analyzed in 25 LI subjects. The total quantity of hydrogen produced was not significantly different during the 6 h after the consumption of Jersey milk or milk containing A2 β-casein only when compared with consumption of conventional milk (Table 3 and Figure 4).

**Figure 3.** Total symptoms reported during the 6 h after consuming the four milk products in 33 lactose maldigesters. \*\* *p* = 0.001 for abdominal pain and \* *p* = 0.04 for total symptoms (abdominal pain + bloating + flatulence + diarrhea) due to milk containing A2 β-casein only vs. conventional milk; † *p* = 0.05 for bloating due to Jersey milk versus conventional milk.

**Figure 4.** Total hydrogen produced during the 6 h after consuming the four milk products in 25 lactose intolerant subjects. ppm, parts per million. † *p* = 0.05, † *p* = 0.03, † *p* = 0.01, and † *p* = 0.05 for Jersey milk vs. commercial milk at 0, 0.5, 2, and 3 h, respectively.

#### 3.3.2. All Maldigesters

Total hydrogen produced by 33 maldigesters following consumption of milk containing A2 β-casein only was significantly lower compared with hydrogen produced by subjects following consumption of conventional milk (13,771 vs. 16,460 ppm; *p* = 0.04). However, hydrogen production following consumption of Jersey milk was not significantly different from that following consumption of conventional milk (15,079 vs. 16,460 ppm; *p* = 0.17) (Table 3 and Figure 5).

**Figure 5.** Total hydrogen produced during the 6 h after consuming the four milk products in 33 lactose maldigesters. ppm, parts per million. \* *p* = 0.05, \* *p* = 0.03 for milk containing A2 β-casein only vs. conventional milk at 3 and 4 h, respectively; † *p* = 0.03 for Jersey milk versus conventional milk at 2 h.

#### *3.4. Adverse Events*

There were no adverse events or unintended harmful effects reported by subjects due to the consumption of the four different types of milk.

#### **4. Discussion**

The results of our study indicate that the consumption of milk containing A2 β-casein only produced fewer GI symptoms in lactose maldigesters compared with consumption of conventional milk. On the other hand, Jersey milk did not reduce GI symptoms, compared with conventional milk. In LI subjects and lactose maldigesters, milk containing only A2 β-casein significantly decreased abdominal pain compared with the consumption of conventional milk. Conversely, the consumption of Jersey milk was not associated with reduced abdominal pain.

The effects of the milk treatments on GI symptoms may be related to GI effects due to longer transit time in the colon by milk containing A1 β-casein [17]. A study in Wistar rats and some human clinical trials showed that A1 β-casein increased GI transit time and colonic activity of the inflammatory marker myeloperoxidase [17,18,30]. These effects, which were counteracted by the opioid blocker naloxone, might be initiated and mediated by the opioid peptide BCM-7, which is formed after the ingestion of A1 β-casein [30]. Further, bovine casein-derived opioid peptides can inhibit cysteine uptake in both GI epithelial and neuronal cells, resulting in elevated oxidative stress and altered DNA methylation, including on genes that are important for mediating inflammation [31].

Consistent with the results reported herein, in Chinese subjects with self-reported LI, consumption of conventional milk (equivalent to the conventional milk tested in the present study) produced more GI symptoms than did milk containing A2 β-casein only [17]. Moreover, increased GI transit times and concentrations of serum inflammatory markers IgG and IL-4 were noted after the consumption of milk containing A1/A2 β-casein rather than milk containing A2 β-casein only. Therefore, the increase in GI symptoms may be due to inflammation and GI transit time, suggesting a need for further investigation.

The presence of LI should be confirmed by recording symptom scores for abdominal pain, bloating, flatulence, and diarrhea [26], and previous studies of A1 and A2 β-caseins in LI individuals did not specifically use QLCSS to screen subjects [15–18]; as a result, these earlier studies selected subjects with

perceived LI for evaluation, many of whom may not have been truly intolerant. In contrast, we verified LI via symptom scores during screening. Thus, our study is the first to demonstrate that verified LI individuals are able to better tolerate a single meal of milk containing only A2 β-casein compared with conventional milk containing both A1 and A2 β-casein.

Because our study included only 25 LI subjects, the results may not be generalizable to larger populations, although this limitation is offset by the racial and ethnic diversity of the study population. Notably, the strict QLCSS inclusion criteria contributed to the small sample size. However, given the greater statistical significance when including the eight maldigesters without symptoms of intolerance, this rigorous inclusion criterion might not be important in the population we studied. Furthermore, the fact that only 35 of 94 eligible milk avoiders/perceived intolerant individuals met the maldigestion criteria suggests that the number of people in our population with perceived or self-reported LI markedly exceeds the number with actual LI verified by symptom scores and maldigestion.

We did not categorize the subjects into age groups. Among the 25 LI study participants, 22 were in the 19- to 35-year-old age group, and three were in the 36- to 50-year-old age group. However, in a study of 600 participants aged 20–50 years who were stratified into two groups (20–35 years and 36–50 years), age had no effect on GI symptoms after milk consumption [15]; nonetheless, adults aged >50 years might respond differently to milk ingestion and this requires further evaluation.

The BMI of subjects in our study ranged from 18–33 kg/m2. Normal-weight, overweight, and obese subjects were included in the study, but severely obese individuals were excluded. The impact of BMI differences on GI symptoms was reduced by providing subjects with calculated quantities of milk with respect to bodyweight, something that was not done in previous studies [15–18].

The effects of A1 β-casein, or its digestive by-product BCM-7, appear to be acute in our study. That is, the effects of a single milk challenge in our study were monitored over a short period (30 min to 6 h), and results may have been different if a multi-meal or multi-day feeding trial had been conducted in the same population. Long-term feeding effects might worsen GI symptoms and prior studies have shown sustained inflammatory effects with A1 β-casein, which could worsen symptoms [17,18,32]. Furthermore, there are also changes in microbial metabolites such as butanoic acid, acetic acid, and propanoic acid in adults and children following consumption of A1 β-casein [17,18], showing that A1 β-casein affects the microbiota in the gut.

#### **5. Conclusions**

In summary, results of the analysis of 25 LI subjects revealed significantly lower abdominal pain after the consumption of milk containing only A2 β-casein compared with the consumption of conventional milk. Total breath hydrogen produced by LI subjects was not significantly different from that after the consumption of conventional milk, possibly because our sample size was too small to detect differences in breath hydrogen production. The reduction in abdominal pain after the consumption of milk containing A2 β-casein only, compared with the consumption of conventional milk, was consistent with the results of another clinical trial [15]. Among the eight maldigesters tested since June 2019, none met the LI criteria during screening or intervention. In the post hoc analysis, symptoms of intolerance were not reduced after the consumption of Jersey milk compared with conventional milk, potentially because of the presence of some A1 β-casein in Jersey milk. However, there was a significant reduction in symptoms among these 25 individuals and eight additional lactose maldigesters following the consumption of milk containing only A2 β-casein. These findings warrant confirmation in larger study populations.

**Author Contributions:** Conceptualization, D.A.S. and O.M.S.; formal analysis, M.R., D.A.S., and O.M.S.; data curation, M.R., O.M.S., and T.K.E.; writing—original draft preparation, M.R.; writing—review and editing, D.A.S. and T.K.E. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was co-funded by the a2 Milk Company Ltd. and National All-Jersey Inc.

**Acknowledgments:** Eurofins conducted the analysis of major nutrients in the four milk treatments; Purdue Proteomics Facility analyzed the A1/A2 β-casein ratio using mass spectrometry. We would also like to thank the Statistics Department at Purdue University, West Lafayette, IN, USA for helping with the power calculation, and staff members in the metabolic kitchen and Clinical Space at Purdue University. We thank David Murdoch, BSc (Hons) of Edanz Evidence Generation for providing editorial support.

**Conflicts of Interest:** The authors declare no potential conflicts of interest during the conduct of this study.

#### **Appendix A**

*Inclusion and Exclusion Criteria for Phone Screening*

Inclusion criteria:

	- a. At least one score of "moderately severe" or "severe" on a single symptom during the 6 h HBT
	- b. A score of "moderate" or greater for a single symptom at least two timepoints during the 6 h HBT
	- c. At least one "moderate" score or greater for each of two symptoms during the 6 h HBT

Exclusion criteria:


#### **References**


32. Yadav, S.; Yadav, N.D.S.; Gheware, A.; Kulshreshtha, A.; Sharma, P.; Singh, V.P. Oral feeding of cow milk containing A1 variant of β casein induces pulmonary inflammation in male Balb/c mice. *Sci. Rep.* **2020**, *10*, 8053. [CrossRef]

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* β**-Lactoglobulin Elevates Insulin and Glucagon Concentrations Compared with Whey Protein—A Randomized Double-Blinded Crossover Trial in Patients with Type Two Diabetes Mellitus**

**Stine B. Smedegaard 1,\*, Maike Mose 1, Adam Hulman 1, Ulla R. Mikkelsen 2, Niels Møller 3, Gregers Wegener 4,5, Niels Jessen <sup>1</sup> and Nikolaj Rittig 1,3**


**Abstract:** Whey protein is an insulinotropic fraction of dairy that reduces postprandial glucose levels in patients with type 2 diabetes mellitus (T2DM). We have recently shown that β-lactoglobulin (BLG), the largest protein fraction of whey, elevates insulin concentrations compared with isonitrogenous whey protein isolate (WPI) in healthy individuals. We therefore hypothesized that BLG pre-meals would lower glucose levels compared with WPI in patients with T2DM. We investigated 16 participants with T2DM using a randomized double-blinded cross-over design with two pre-meal interventions, (i) 25 g BLG and (ii) 25 g WPI prior to an oral glucose tolerance test (OGTT), followed by four days of continuous glucose monitoring (CGM) at home. BLG increased concentrations of insulin with 10%, glucagon with 20%, and glucose with 10% compared with WPI after the OGTT (all *p* < 0.05). Both BLG and WPI reduced the interstitial fluid (ISF) glucose concentrations (using CGM) with 2 mM and lowered glycemic variability with 10–15%, compared with tap-water (*p* < 0.05), and WPI lowered the ISF glucose with 0.5 mM compared with BLG from 120 min and onwards (*p* < 0.05). In conclusion, BLG pre-meals resulted in higher insulin, glucagon, and glucose concentrations compared with WPI in participants with T2DM. Pre-meal servings of WPI remains the most potent protein in terms of lowering postprandial glucose excursions.

**Keywords:** type 2 diabetes mellitus; whey; glucose; glycemic variability; beta-lactoglobulin; pre-meal; CGM

#### **1. Introduction**

Pre-meals of whey protein have shown promising effects on the subsequent glucose trajectories in both healthy participants and patients with type 2 diabetes mellitus (T2DM) [1,2]. Whey given 15–30 min before a meal mediates a rise in insulin concentration and results in lower postprandial blood glucose concentrations [1,3]. The underlying mechanisms behind the insulinotropic properties observed following whey protein consumption are complex and not fully understood. Whey is especially rich in the branched chained amino acid (BCCA), leucine, which has direct insulin stimulating effect on the beta cell of the pancreas [4]. Whey protein also increases the concentration of the incretin hormones glucose-dependent insulinotropic polypeptide (GIP) [1,5,6] and glucagon-like peptide-1 (GLP-1) [1,5,7], which are also known to stimulate insulin secretion. Data from mouse

**Citation:** Smedegaard, S.B.; Mose, M.; Hulman, A.; Mikkelsen, U.R.; Møller, N.; Wegener, G.; Jessen, N.; Rittig, N. β-Lactoglobulin Elevates Insulin and Glucagon Concentrations Compared with Whey Protein—A Randomized Double-Blinded Crossover Trial in Patients with Type Two Diabetes Mellitus. *Nutrients* **2021**, *13*, 308. https://doi.org/10.3390/nu13020308

Received: 22 December 2020 Accepted: 20 January 2021 Published: 22 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

pancreatic islets suggests that the exposure to an amino acid mixture and GIP [4], rather than one specific amino acid, has the greatest insulinotropic effects on the beta-cell.

Milk protein consists of around 80% casein and 20% whey [8]. Whey protein consists of 50–60% β-lactoglobulin (BLG), 17% α-lactalbumin, 10% immunoglobulins, 5% albumin and other polypeptides [9]. Recent data from our group show that BLG increases the serum(s)-concentration of insulin 23% more than a regular iso-nitrogenous whey protein isolate (WPI) in individuals without prior health issues. This observation led us to the hypothesis that pre-meal servings of BLG would stimulate insulin secretion and lower glucose trajectories compared with WPI in patients with T2DM. A more potent protein may lower protein and excessive calorie intake and improve compliance in prolonged protein pre-meal treatment regimes. Therefore, we performed a randomized double-blinded crossover trial to investigate the effects of BLG and WPI pre-meals in patients with T2DM.

#### **2. Materials and Methods**

#### *2.1. Study Approval*

The study complied with the Declaration of Helsinki and was approved by the regional research ethics committee (1-10-72-226-19), registered at ClinicalTrials.gov (NCT04166760), and applied to the regulations of the Danish Data Protection Agency. All participants gave their written informed consent before inclusion in the study.

#### *2.2. Participants*

Participants were eligible for inclusion if they had T2DM, were between 18 and 80 years old, had a BMI between 20 and 35 kg/m2, had hemoglobin (Hb)-A1c between 40 and 69 mmol/L, and C-peptide between 370 and 1200 pmol/L. Recruitment was performed through social media (Facebook) and local newspapers. Exclusion criteria were milk allergies, daily intake of protein supplements, anti-glycemic medication other than metformin, or inability to speak or understand Danish. All participants were screened with a blood test panel of HbA1c, creatinine, thyrotropin, C-reactive-protein, sodium, potassium, albumin, alanine aminotransferase, alkaline phosphatase, bilirubin, hemoglobin, and C-peptide before inclusion.

#### *2.3. Design and Protocol*

The study was a randomized, double-blinded, cross-over trial with two interventions. Study days were identical except for interventions and consisted of an oral glucose tolerance test (OGTT) performed in our laboratory and four days of monitoring at home. The study was performed at the Steno/Medical laboratory, Aarhus University Hospital, Denmark. The two interventions consisted of: (i) BLG and (ii) whey protein isolate (WPI). There was a minimum washout period of one week and a maximum of six weeks between the two OGTTs. Participants and investigators were blinded in regard to the interventions. For an overview of the design and randomization, see Figure 1.

Before attending the laboratory, participants were asked to eat according to Danish nutritional guidelines (15% fat, 30% protein, and 55% carbohydrates) for 48 h and to avoid strenuous physical activity before and during each of the investigations (laboratory and home monitoring). If participants received metformin, this treatment was discontinued for five days before and during the investigations. All participants arrived following a 10-h overnight fast. During each study day, an intravenous catheter was placed in an antecubital vein for blood sampling. The participants consumed either 25 g of WPI or BLG 30 min before a 75 g OGTT was performed. Blood samples were collected consecutively in the three following hours.

Following the OGTT investigation, participants were equipped with a continuous glucose monitor (CGM), an activity monitor, four standardized breakfasts, a protein drink shaker, and four small plastic bags with 25 grams of the protein intervention. The first 24 h of CGM and activity recordings were used to calibrate equipment and excluded from analyses. Each participant was randomized to consume the protein pre-meals 30 min before the standardized breakfast and dinner during days two and three or during days four and five (Figure 1). Participants consumed an iso-voluminous amount of tap-water (CTR) 30 min before breakfast and dinner during the days without protein pre-meals. They were asked to avoid strenuous physical activity and eat similarly during the days of home-monitoring. The participants filled out a food-diary with timestamps for pre-meals and meals to ensure compliance and perform CGM analyses.

**Figure 1.** Flowchart of the study investigations. Participants were randomized to consume one of two pre-meals, (i) βlactoglobulin (BLG) or (ii) whey protein isolate (WPI), 30 min before a 75 g oral glucose tolerance test (OGTT) in our laboratory or before breakfast and dinner at home. Participants were equipped with a continuous glucose monitor (CGM), an activity monitor, and standardized breakfast meals. Participants were also randomized to consume pre-meals before breakfast and dinner on days 2–3 or days 4–5 and control (tap-water) on the other two days. The experiment was repeated after 1 to 6 weeks from the OGTT.

#### *2.4. Interventions and Meals*

The primary investigator enrolled and assigned participants to the sequence of interventions using www.randomizer.org [10]. The WPI (Lacprodan DI-9213) and BLG were provided by Arla Foods Ingredients Group P/S, Viby J, Denmark. The interventions were similar in appearance and taste. Two persons without relation to the investigations dosed and blinded 25 g of protein in small, labeled plastic bags. The proteins were dissolved in 200 mL of tap-water and served as a shake. The standardized breakfast consisted of 50 g cornflakes (Vores Cornflakes 500 g), 31 g raisins (Svansoe Rosiner 1500 g), and 250 mL skimmed milk (Arla® Skummetmaelk 0.1% 250 mL) equivalent to 77.6 g carbohydrates, 13.9 g protein, 1 g fat/375 kCal. The characteristics of WPI and BLG are shown in Table 1. Participants and all persons involved in the trials, including the outcome assessors, remained blinded until statistical analyses had been performed. There were no adverse events reported.

**Table 1.** Composition of the interventions; β-lactoglobulin (BLG) and whey protein isolate (WPI).



**Table 1.** *Cont.*

The composition of amino acid analysis was done by Eurofins (GLP (Good Laboratory Practice) certified), and the nutritional content analysis was done by Danmark Protein (Arla Foods Ingredients).

#### *2.5. Blood Analysis*

Blood samples were drawn at −30, 0, 10, 20, 30, 40, 50, 60, 90, 120, 150, and 180 min following the OGTT. Plasma(p)-glucose was measured immediately using YSI 2300 model Stat Plus glucose analyzer (YSI Incorporated, Yellow Springs, OH). Blood for the remaining analyses was centrifuged at 4 ◦C, frozen at −20 ◦C, stored at −80 ◦C, and analyzed on the same assay after both arms of the study were completed for all participants. Sinsulin, s-C-peptide, and p-glucagon concentrations were measured with an enzyme-linked immunosorbent assay (ELISA) technique using a commercial kit (Mercodia Insulin ELISA, Mercodia Glucagon ELISA, Mercodia C-peptide ELISA, Sweden). S-free fatty acids (FFA) were measured using the in vitro enzymatic colorimetric method assay NEFA-HR(2), which quantifies the concentration of non-esterified fatty acids (FUJIFILM Wako Chemicals Europe GmbH, Germany). P-amino acids (AA) concentrations were measured by high-pressure liquid chromatography (HPLC) method using a Thermo Scientific Ultimate 3000 system, as earlier described [11]. Briefly, the samples were diluted 1.11x by adding 2 M Perchloric acid (HCIO4) and then centrifuged at 14,000× *g* at 4 ◦C for 10 min. The supernatant was removed and filtered through a spin filter (0.22 μm) at 14,000× *g* for 1 min; then diluted 50× with 0.2 M HCIO4 to a final dilution factor of 55.5. Hereafter, the samples were injected into the HPLC. For separation, a Kinetex EVO C18 2.6 μm 4.6 × 150 mm column from Phenomenex, U.S., was used. Detection was done by fluorometric detection with excitation on 337 nm and emission on 442 nm. Samples for p-GIP and p-GLP-1 were extracted in final concentrations of 70% ethanol before analyses. Samples were analyzed on radioimmunoassays using antiserum #89390 for GLP-1 and antiserum #80867 for GIP targeting the C-terminal end of the hormones reacting equally with the intact hormone and the primary metabolites (N-terminally truncated) [12,13].

#### *2.6. Continuous Glucose Monitoring*

Continuous measurement of glucose concentrations in the interstitial fluid (ISF) was performed using a CGM device (NordicInfu Care Denmark, Dexcom G6, Dexcom Inc., San Diego, CA, USA). The device measures glucose every five minutes via a subcutaneous sensor. The participants wore the device on the abdomen and were unaware of their glucose level as the receiver was blinded. Data were uploaded to and analyzed in the software CLARITY (Dexcom CLARITY, v3.32.0, Dexcom Inc., San Diego, CA, USA). The mean glucose ± standard deviation (SD), daily maximum glucose level, and the coefficient of variation (CV) was used as outcome measures.

#### *2.7. Activity Monitoring*

A combined accelerometer and heart rate (HR) monitor (Actiheart 5 (AH), CamNtech Limited, Cambridge, UK) was used to evaluate the activity and estimate energy expenditure during the home investigation period. The AH unit was worn on and connected to the chest with two self-adhesive electrodes—one below the sternum and one under the left pectoral muscle. Data on accelerometry was collected at 32 Hz, and HR was collected as inter-beatintervals. Data from the unit was uploaded to and analyzed using AH software (Actiheart software, version 5.1.10, camNtech ltd., Cambridge, UK). The software, processing of data, and validation of the system have been described in detail elsewhere [14]. Briefly, the AH software has a built-in function to correct missing beats and clean noise from the HR data. The software uses the cleaned HR data and data on activity in the integrated branched chained model "Group Cal JAP 2007" to estimate activity energy expenditure (AEE). In the case of missing HR data >5 min, the AEE is solely based on activity. The software provides the total energy expenditure (TEE) from a model using weight, height, age, sleeping heart rate (resting heart rate—10), AEE, and diet-induced thermogenesis. The software provides variables on HR, maximum HR, activity counts, AEE, and TEE.

#### *2.8. Statistical Analysis*

Statistical analyses and figures were conducted using the nlme (version 3.1-142), Epi (version 2.37) packages in R (R Foundation for Statistical Computing, Vienna, Austria, version 3.6.2) and SigmaPlot (San Jose, CA, USA, version 14.0). Trajectories on substrates and hormones in relation to the OGTT were fitted using random-effects models with a natural cubic spline specification for time. The number and position of the knots are different for those outcomes measured at a different set of time points. Interaction terms were included for each time term and a binary variable coding the two interventions. This, in combination with appropriate contrast matrices, allowed us to estimate trajectories for both interventions and their difference at any time point during the investigation. The differences between trajectories were expressed as percentages, as the outcomes were log-transformed (natural logarithm) before running the models due to their skewed distributions. Individual specific random intercepts and slopes were included in the models to account for the dependence within the data due to its repeated measurement nature. The same method was used to assess glucose trajectories during the three hours following breakfast and dinner for BLG, WPI, and CTR. For this analysis, measurements were included if their time points were after, but within three hours of, the recorded time of breakfast and dinner. The incremental area under the curve (iAUC) was calculated using the trapezoidal approach [15], and a paired t-test or one-way RM ANOVA was used for comparison of each outcome.

CGM-based summary measures and activity characteristics were compared between interventions and controls using random effects models with individual specific random intercepts to account for the cross-over design of the experiment. Differences between groups and their 95% confidence intervals (CIs) were estimated using the appropriate contrast matrices.

A pre-study power calculation with a significance level of 0.05 and a power of 80% was performed. We expected to detect a 25% difference in iAUC in insulin concentration (which was the primary outcome) between BLG and WPI with a 23% SD following the OGTT. This resulted in a sample size of 14. We expected a dropout rate of 10% and therefore included 16 participants.

#### **3. Results**

#### *3.1. Participants*

Sixty-five individuals were initially screened by the primary investigator over the phone (Figure 2). Sixteen participants were included and completed the studies between January 2020 and June 2020. One participant was unable to complete the home monitoring program. Patient characteristics are shown in (Table 2). There was a median washout-period of 9 days (range 7–23 days) between laboratory investigations.

**Figure 2.** Flow diagram of inclusion in the randomized cross-over trial. WPI, whey protein isolate; BLG, β-lactoglobulin.

**Table 2.** Demographic characteristics.


Data are presented as absolute numbers or medians (ranges).

#### *3.2. Oral Glucose Tolerance Test (OGTT)*

#### 3.2.1. Substrate and Hormone Concentrations

The p-glucose concentration was higher 120 min following BLG compared with WPI ingestion and reached a maximum difference of 10% 180 min following the OGTT (Figure 3A). The s-insulin concentration was elevated with 10% at 30–60 min and pglucagon with 20% at 60–90 min after the OGTT following BLG compared with WPI (Figure 3B,C). Both BLG and WPI elevated s-C-peptide concentrations with no difference between BLG and WPI (Figure 3D). The WPI led to a higher insulin/glucagon ratio at 60 min (Figure 4A). Both proteins elevated p-GIP and p-GLP-1 concentrations and suppressed s-FFA concentrations with no difference between BLG and WPI (Figure 4B–D).

**Figure 3.** Plasma (p) and serum (s) concentrations of hormones and substrates after β-lactoglobulin (BLG) and whey protein isolate (WPI) pre-meals 30 min before an OGTT (0 min). Panels to the left show trajectories of the mean concentration (solid lines) with 95% confidence intervals (95% CIs) (dashed lines) of (**A**) p-glucose, (**B**) s-insulin, (**C**) p-glucagon, (**D**) s-C-peptide after WPI (blue) and BLG (red) consumption. The mean relative difference (solid line, purple) with 95% CIs (dashed lines) between the two interventions is shown in the middle panels. Panels to the right show the individual incremental area under the curve (iAUC) with a bar plot showing the mean ± standard deviation after WPI (blue) and BLG (red) consumption. *n* = 16.

**Figure 4.** Plasma (p) and serum (s) concentrations of hormones and substrates after β-lactoglobulin (BLG) and whey protein isolate (WPI) pre-meals 30 min before an OGTT (0 min). Panels to the left show trajectories of the mean concentration (solid lines) with 95% confidence intervals (95% CIs) (dashed lines) of (**A**) s-insulin/p-glucagon ratio, (**B**) p-glucose-dependent insulinotropic polypeptide (GIP), (**C**) p-glucagon-like peptide-1 (GLP-1), (**D**) s-free fatty acids (FFA) after WPI (blue) and BLG (red) consumption. The mean relative difference (solid line, purple) with 95% CIs (dashed lines) between the two interventions is shown in the middle panels. Panels to the right show the individual incremental area under the curve (iAUC) with a bar plot showing the mean ± standard deviation after WPI (blue) and BLG (red) consumption. *n* = 16.

#### 3.2.2. Amino Acids

BLG elevated the p-concentration of aspartate, glutamate, leucine, lysine, methionine, phenylalanine, proline, and tyrosine compared with WPI (Figure S1). WPI elevated the pconcentration of glycine, isoleucine, serine, and threonine compared with BLG (Figure S1).

#### *3.3. Home-Monitoring with Continuous Glucose Monitoring (CGM)* 3.3.1. CGM Glucose Trajectories Following Breakfast

Both protein pre-meals lowered postprandial ISF-glucose concentration following breakfast with the largest difference of 15% (WPI) and 17% (BLG) (2 mM) around 60 min compared with CTR (Figure 5). In alignment with our results from the OGTT, the ISFglucose was 7% (0.5 mM) lower after 150 min following WPI compared with BLG, but 4% higher around breakfast consumption (Figure 5).

**Figure 5.** Interstitial fluid concentration of glucose (ISF-glucose) in the 180 min following intake of the pre-meals βlactoglobulin (BLG), whey protein isolate (WPI), and control: tap-water (CTR) (−30 min) and standardized breakfast at home (0 min). The panel to the left shows trajectories of the mean ISF-glucose (solid lines) with 95% confidence intervals (95% CIs) (dashed lines) after WPI (blue), BLG (red) and CTR (yellow) consumption. The mean relative differences (solid lines) with 95% CIs (dashed lines) between the WPI and CTR (blue), BLG and CTR (red) and WPI and BLG (purple) are shown in the middle panel. The panel to the right shows the individual incremental area under the curve (iAUC) with a bar plot showing the mean ± standard deviation after WPI (blue), BLG (red) and CTR (yellow) consumption. One-way repeated measure ANOVA, *p* = 0.002, and post hoc (Student–Newman–Keuls) paired *t*-tests: a, WPI vs. CTR: *p* = 0.002; b, BLG vs. CTR: *p* = 0.077; c, WPI vs. BLG: *p* = 0.052. *n* = 15.

#### 3.3.2. CGM and Summary Statistics

There was no difference in mean ISF-glucose between BLG, WPI, and CTR. The glycemic variability expressed as the CV was lower by 10% during WPI and by 15% during BLG, and the SD was lower by 9% during WPI and by 13% during BLG, compared with CTR. Additionally, after breakfast, the maximum glucose concentration was lower by 13% during WPI and 12% during BLG compared with CTR. The daily maximum glucose level was lower by 7% during WPI and 5% during BLG compared with CTR. No statistically significant differences were detected between BLG and WPI in any of the CGM summary variables (Table 3).

#### *3.4. Energy Expenditure*

TEE and AEE were higher on days with BLG compared with WPI. There was no significant difference between days with protein compared with CTR. Participants had similar activity counts, HR, and maximum HR on days with protein and days with CTR (Table 4).


**Table 3.** Summary variables on continuous glucose monitoring.

Coefficient of variation (CV) and standard deviation (SD) as parameters on glycemic variability. Maximum (max) after breakfast and dinner is the peak in the postprandial glucose concentration during three hours following the meals. Values are presented as means with 95% confidence intervals. \* *p* < 0.05, \*\* *p* < 0.01, statistically significant differences are highlighted (bold). CTR, control; WPI, whey protein isolate; BLG, beta-lactoglobulin.

**Table 4.** Activity measurements and energy expenditure.


Data are expressed as means with 95% confidence intervals. \* *p* < 0.05, statistically significant differences are highlighted (bold). TEE, total energy expenditure; kCal, kilocalories; AEE, activity energy expenditure; HR, heart rate; BPM, beats per minute; CTR, control; WPI, whey protein isolate; BLG, beta-lactoglobulin.

#### **4. Discussion**

In this study, we showed how a BLG pre-meal served 30 min before an OGTT resulted in higher concentrations of insulin, glucagon, and glucose compared with WPI in patients with T2DM. The study was originally designed to investigate the insulinotropic properties of BLG with the hypothesis that elevated insulin concentrations would lower postprandial glucose excursions compared with WPI. We confirmed that BLG elevates insulin concentrations compared with WPI, but the simultaneous glucagonotropic effect also associated with BLG most likely explains why ISF-glucose concentrations were slightly higher (0.5 mM) following BLG compared with WPI, opposing our original hypothesis. Despite the similarity between the two dairy products, BLG contained more leucine and phenylalanine than WPI, which was also present in the p-concentrations of these specific AA following interventions. Both leucine and phenylalanine have been shown to stimulate insulin secretion [4,16] which, to some extent, may explain the insulinotropic properties. Also, p-concentrations of methionine and tyrosine have been shown to correlate with glucagon concentrations in humans [17], and perfusion studies in dogs and rodents have shown that aspartate, glutamate, lysine, and proline stimulate glucagon secretion [18,19]. These AA were all significantly higher after BLG consumption compared with WPI. Glucagon release is potently stimulated by GIP [20], but plasma concentrations of GIP following interventions were comparable between interventions. Hence, we suggest that the insulinotropic and glucagonotropic effects associated with BLG may relate to its specific AA composition.

Our study is the first to show glucose-lowering effect of pre-meal whey protein in a home-setting using CGM. Both interventions lowered glucose excursions with 2 mM following a standardized breakfast compared with tap-water (CTR). The effect is in line with other studies investigating similar doses of whey protein pre-meal servings [6,21]. To our knowledge, only one other study has investigated whey pre-meals in individuals with T2DM using CGM in a home-setting [22]. This study compared whey protein with a mixture of indigestible potato starch (carbohydrate rich) and could not show statistically significant effects on mean glucose levels, glucose trajectories, or glycemic variability following meals. It is well known that ingestion of small amounts of carbohydrates preceding a glucose load lowers the following glucose excursion, an effect referred to as the Staub-Traugott effect [23,24] that may explain why no significant differences were found in this study.

Only one study on long-term pre-meal whey exposure in T2DM has been performed (12 weeks) [21]. This study showed a small significant reduction in HbA1c (−1 mmol/mol). It should be noted that the participants were already well regulated with an HbA1c of 49 mmol/mol, which may have affected the size of the outcome. However, HbA1c does not necessarily reflect postprandial glucose excursions [25], and results might have been more substantial on glycemic variability. Large glucose excursions and high glycemic variability have been associated with risk of cardiovascular disease [26] as well as impaired cognitive function [27]. This emphasizes that minimizing postprandial glucose excursions may be important in the management of T2DM. We showed a reduction in glycemic variability, maximum glucose levels, and lower glucose excursions after consuming the pre-meal proteins compared with tap-water. Future long-term studies on pre-meal whey protein in participants with T2DM should preferably include investigations on glycemic variability.

Our study was limited, as the CTR intervention (tap-water) was unblinded. However, morning glucose concentrations were comparable between conditions, the participants were given the same standardized breakfast meal, and activity levels were similar between days with protein interventions and CTR. Activity and energy expenditure were comparable between groups, but showed a minor statistically significant elevation in TEE and AEE during BLG compared with WPI. These differences were small and only strengthen our findings showing lower glucose levels during WPI compared with BLG. We instructed participants to eat according to the Danish national recommendations (55% carbohydrates, 30% fat, and 15% protein) and to eat similar portion sizes during the home monitoring period. Still, the food diaries were generally of poor quality and lacked information. We did not include a control condition (e.g., tap-water) in the OGTT experiment because the primary aim of the study was to compare BLG and WPI. Pre-meals of whey have, in many previous studies, already proven effective in lowering glucose concentrations [2,3,21,28], but direct comparisons to other proteins are sparse.

A major strength of our study is the combination of investigations in a controlled laboratory and home setting. We included both men and women in our trial and both CGM and activity monitors in our investigations. The cross-over design eliminated any inter-individual differences.

In conclusion, a pre-meal of BLG elevates insulin, glucagon, and glucose concentration compared with WPI following an OGTT in patients with T2DM. Both WPI and BLG lowered glycemic variability and glucose trajectories compared with tap-water. WPI remains the most potent pre-meal in the management of postprandial glucose excursions.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-6 643/13/2/308/s1, Figure S1: Plasma concentrations amino acids (AA) after β-lactoglobulin (BLG) and whey protein isolate (WPI) pre-meals 30 min before an OGTT (0 min). Panels to the left show trajectories of the mean concentration (solid lines) with 95% confidence intervals (95%CIs) (dashed lines) of (A) alanine, (B) arginine, (C) aspartate, (D) glutamate, (E) glycine, (F) histidine, (G) isoleucine, (H) leucine, (I) lysine, (J) methionine, (K) phenylalanine, (L) proline, (M) serine, (N) threonine, (O) tyrosine, (P) valine, (Q) total AA after WPI (blue) and BLG (red) consumption. The mean relative difference (solid line, purple) with 95%CIs (dashed lines) between the two interventions is shown in the middle panel. Panels to the right show the individual incremental area under the curve (iAUC) with a bar plot showing the mean ± standard deviation after WPI (blue) and BLG (red) consumption. N = 16.

**Author Contributions:** Conceptualization, M.M., U.R.M., N.J., and N.R.; data curation, S.B.S.; formal analysis, S.B.S. and A.H.; funding acquisition, M.M. and N.R.; investigation, S.B.S.; project administration, S.B.S.; resources, N.M., G.W., and N.J.; software, A.H.; supervision, N.R.; visualization, S.B.S. and A.H.; writing—original draft, S.B.S.; writing—review and editing, N.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Arla Food Ingredients Group P/S, Soenderhoej 10-12, 8260 DK-Viby, CVR 33372116.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of The Central Denmark Region Committees on Health Research Ethics (1-10-72-226-19, November 5, 2019).

**Informed Consent Statement:** Informed consent was obtained from all participants involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy.

**Acknowledgments:** We are grateful for the outstanding technical support performed by Elsebeth Hornemann, Hanne Fjeldsted Petersen, Lene Albaek, Lisa Buus and Per Fuglsang Mikkelsen. In addition, we would like to thank Arla Foods Ingredients P/S for providing the protein interventions and funding the project. Last, we thank all participants for their participation in the study.

**Conflicts of Interest:** The study was supported by Arla Foods Ingredients Group P/S. U.R.M. is employed as a nutrition research scientist at Arla Foods Ingredients Group P/S. Arla Foods amba, Soenderhoej 14, 8260 Viby J, Denmark, has a pending patent application regarding BLG, with N.R., M.M., N.M., A.H., U.R.M., and S.B.S. as co-inventors. N.R., M.M., N.J. and U.R.M. took initiative to and designed the study together. U.R.M. and Arla Food Ingredients P/S had no role in the collection, analyses, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Article* **Bovine Milk-Derived Emulsifiers Increase Triglyceride Absorption in Newborn Formula-Fed Pigs**

**Kristine Bach Korsholm Knudsen 1,2, Christine Heerup 3, Tine Røngaard Stange Jensen 1, Xiaolu Geng 4, Nikolaj Drachmann 5, Pernille Nordby 6, Palle Bekker Jeppesen 7, Inge Ifaoui 2, Anette Müllertz 3, Per Torp Sangild 1, Marie Stampe Ostenfeld <sup>5</sup> and Thomas Thymann 1,\***


**Abstract:** Efficient lipid digestion in formula-fed infants is required to ensure the availability of fatty acids for normal organ development. Previous studies suggest that the efficiency of lipid digestion may depend on whether lipids are emulsified with soy lecithin or fractions derived from bovine milk. This study, therefore, aimed to determine whether emulsification with bovine milk-derived emulsifiers or soy lecithin (SL) influenced lipid digestion in vitro and in vivo. Lipid digestibility was determined in vitro in oil-in-water emulsions using four different milk-derived emulsifiers or SL, and the ultrastructural appearance of the emulsions was assessed using electron microscopy. Subsequently, selected emulsions were added to a base diet and fed to preterm neonatal piglets. Initially, preterm pigs equipped with an ileostomy were fed experimental formulas for seven days and stoma output was collected quantitatively. Next, lipid absorption kinetics was studied in preterm pigs given pure emulsions. Finally, complete formulas with different emulsions were fed for four days, and the post-bolus plasma triglyceride level was determined. Milk-derived emulsifiers (containing protein and phospholipids from milk fat globule membranes and extracellular vesicles) showed increased effects on fat digestion compared to SL in an in vitro digestion model. Further, milk-derived emulsifiers significantly increased the digestion of triglyceride in the preterm piglet model compared with SL. Ultra-structural images indicated a more regular and smooth surface of fat droplets emulsified with milk-derived emulsifiers relative to SL. We conclude that, relative to SL, milk-derived emulsifiers lead to a different surface ultrastructure on the lipid droplets, and increase lipid digestion.

**Keywords:** preterm neonates; fat; gastric lipase; absorption; intestine; milk; emulsions; vegetable oil; soy lecithin

#### **1. Introduction**

Exclusive breastfeeding is, according to WHO, the preferred nutrition from birth to the age of six months [1], yet worldwide this is only accomplished for approximately 41% of all

**Citation:** Bach Korsholm Knudsen, K.; Heerup, C.; Røngaard Stange Jensen, T.; Geng, X.; Drachmann, N.; Nordby, P.; Bekker Jeppesen, P.; Ifaoui, I.; Müllertz, A.; Torp Sangild, P.; et al. Bovine Milk-Derived Emulsifiers Increase Triglyceride Absorption in Newborn Formula-Fed Pigs. *Nutrients* **2021**, *13*, 410. https:// doi.org/10.3390/nu13020410

Academic Editor: Julianne H. Grose Received: 17 December 2020 Accepted: 22 January 2021 Published: 28 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

infants [2]. Reasons for this include mothers who are unable or chose not to breastfeed or have complications such as preterm birth [3]. Whereas donor milk may be an alternative, particularly for preterm infants [4], high-quality formulas are required when donor milk is not available to ensure survival and normal development. Relative to breastfed infants, formula-fed infants have an increased risk of developing atopic diseases [5], respiratory infections [6], necrotizing enterocolitis [7], and other gastrointestinal complications [8]. There is also evidence of a higher risk of reduced neurodevelopment in formula-fed infants than in infants fed mothers' milk, even after adjusting for important confounders [9,10]. These effects on cognitive outcomes have been observed in both preterm and term infants [9], where formula-fed infants have lower IQ and a lower score for cognitive functions [10,11], which may persist into later life [12,13].

Lipid supplementation to the brain is essential for normal neurological development. Accumulation of lipids in the brain begins in the third trimester and continues the first two years of postnatal life [14]. Especially long-chain polyunsaturated fatty acids (LC-PUFAs) are important, as they represent essential cell membrane components in the brain. A lower concentration of LC-PUFAs in the brain has been observed in formula-fed infants and they also have poorer neurological outcomes [15–17]. There are differences between breast milk and infant formula with regard to the bioavailability of LC-PUFAs, which may partly explain the differences observed in neurodevelopment [18]. However, compensating for lower bioavailability by supplementing infant formula with more LC-PUFAs has not shown any cognitive improvement [19,20].

One reason for the lack of improved neurological outcomes from formulas high in LC-PUFA may be low intestinal absorption leading to low delivery of LC-PUFA to the brain. Fat absorption in infants is generally less efficient relative to adults, and this is even more pronounced in preterm infants who have an absorption rate of 70–80% relative to 95% in adults. This is mainly due to the involvement of different lipases in infants than in adults and thereby different digestive capacity [21]. Moreover, formula-fed infants have reduced fat digestion and absorption relative to infants fed mother's milk [22,23]. Accordingly, lipid digestion in formula-fed infants should be improved to approximate breastfed infants' absorption levels. In mother´s milk, lipids in the form of triglycerides, are mainly transported as milk fat globules surrounded by a milk fat globule membrane (MFGM) [24]. Milk fat globules are secreted into the milk from the mammary gland by a unique mechanism giving MFGM a triple-layered membrane. This is a highly complex membrane containing several classes of phospholipids (including sphingomyelin), and glycosphingolipids, cholesterol, and unique membrane proteins, many of which are highly glycosylated [25]. MFGM forms a hydrophilic layer around the triglyceride core, making the fat globules water-soluble in the milk's water matrix [26,27]. Another phospholipidrich source in milk is extracellular vesicles (EVs) [28]. EVs are also secreted from the mammary gland and consist of a lipid bilayer membrane comprised of phospholipids, glycerosphingolipids, cholesterol, and membrane proteins, but are devoid of a central triglyceride core. The EV membrane is rich in sphingomyelin and cholesterol as well as tetraspanins CD9, CD63, and CD81.

In infant formula today, soy lecithin (SL) is commonly used as an emulsifier and stabilizer. SL mainly consists of phosphatidylcholine and does not contain sphingomyelin or cholesterol, both of which play a major role in forming the lipid rafts found in MFGM [27]. Thus SL provides a different surface structure on the lipid droplets in infant formula [29]. Moreover, human milk lipids are more readily digested in preterm infants compared with infant formula [23]. Lipid digestion in infants has been examined using an in vitro model, simulating infant gastrointestinal conditions. This model used human gastric aspirate as a source of gastric lipase and porcine pancreatin as a source of pancreatic lipase. A higher in vitro gastric lipolysis rate was found when emulsifying lipids with milk phospholipids relative to SL, and this was also the case for intestinal lipolysis rate [30].

On this background, we hypothesized that emulsification with bovine phospholipid sources such as whey protein concentrate enriched in phospholipids (WPC-PL) or WPC

from acid whey-enriched in extracellular vesicles (WPC-A-EV) with their unique composition of glycolipids, phospholipids, sphingomyelin, and glycosylated proteins joined in a complex membrane structure, would enhance digestion of dietary lipids compared to SL. Accordingly, the objective was to emulsify vegetable-based oils with either SL or WPC-PL or WPC-A-EV and determine their effect on the in vitro rate of lipolysis, and in vivo digestion and absorption of triglycerides. To determine any influence of emulsifiers on lipid digestion, we chose to use a preterm neonatal piglet model. This was from the assumption that prematurity per se would associate with lower digestive capacity, thereby making any potential improvements of lipid digestion more clear.

#### **2. Materials and Methods**

#### *2.1. In Vitro Lipolysis*

For the in vitro lipolysis studies, five oil-in-water emulsions were made using either SL (AAK, Karlshamn, Sweden), or bovine whey protein concentrate enriched in phospholipids (WPC-PL), whey protein concentrate from acid whey enriched in triglycerides (WPC-A-TAG), whey protein concentrate from acid whey enriched in EVs (WPC-A-EV), or whey protein concentrate from acid whey enriched in soluble whey protein (WPC-A-WP). All bovine products were kindly donated by Arla Foods Ingredients Group P/S, and the composition of the emulsifiers is provided in Table 1.


**Table 1.** Emulsifier composition of protein, neutral fat, and phospholipids.

PL: phospholipids; PC: phosphatidylcholine; PE: phosphatidylethanolamine; PI: phosphatidylinositol; PS-Na: phosphatidylserine-sodium; SM: sphingomyelin, N/A: not assessed; WPC-A-EV: whey protein concentrate from acid whey-enriched in extracellular vesicles (WPC-A-EV), WPC enriched in phospholipids; WPC-A-TAG: whey protein concentrate from acid whey enriched in triglycerides; WPC-A-WP: whey protein concentrate from acid whey enriched in soluble whey protein; SL: soy lecithin.

> The emulsions were prepared using the emulsification method, composition, and oil-blend described in Heerup et al. (submitted) [31]. In brief, the emulsions were made with 0.35% emulsifier and 3.5% oil-blend (98.92% Akonino NS (AAK, Karlshamn, Sweden), 0.97% MEG-3 (DSM, Mulgrave, NS, Canada), and 0.51% Arasco (DSM)) in an aqueous 11.5 mM CaCl2 and 8.5 mM NaCl solution.

> The emulsions were digested using the in vitro pediatric gastro-intestinal digestion model described in Heerup et al. In brief, the model consisted of a 50 min gastric step at pH 6.4 with 3.75 TBU/mL recombinant human gastric lipase (rHGL) kindly donated by Bioneer A/S (Hørsholm, Denmark) and 126 U/mL pepsin purchased from Sigma Aldrich (St. Louis, MO, USA), a 90 min intestinal step at pH 6.5 with 26.5 TBU/mL pancreatin (Sigma Aldrich), and a back titration to pH 9. Since the SL emulsion was not stable in the 11.5 mM CaCl2 and 8.5 mM NaCl solution, CaCl2 and NaCl were instead added as part of the gastric medium. Table 2 shows the final gastric and intestinal assay composition, including the contribution of CaCl2 and NaCl from the emulsions. The degree of lipolysis over time was measured indirectly by continuous titration of ionized free fatty acids with

0.2 mM NaOH using a Metrohm Titrando pH Stat (Metrohm, Glostrup, Denmark). The particle size distribution of the undigested emulsions was measured on the day of lipolysis.


**Table 2.** Final in vitro gastric- and intestinal digestion assay concentrations (mM). The contribution of CaCl2 and NaCl from the emulsions is included in the shown concentrations.

<sup>1</sup> NaCl was purchased from VWR (Darmstadt, Germany), Tris from ICN Biomedicals (Santa Ana, CA, USA), phospholipids (phosphatidylcholine) from Lipoid (Köln, Germany), maleic acid, and sodium taurocholate from Sigma Aldrich.

#### *2.2. Microstructure of Emulsions*

To determine structural differences we selected a subfraction of emulsions, i.e., SL, WPC-PL, and WPC-A-EV. These emulsions were visualized with transmission electron microscopy (TEM) and cryo-scanning electron microscopy (SEM). The emulsions used for TEM and SEM were prepared according to the method described by Heerup et al., although with no CaCl2 or NaCl added to any of the emulsions. The emulsions were mixed 1:1 with 2% agarose (Carl Roth, Germany) at 37 ◦C and left at room temperature for solidification. Several small pieces of the solidified sample were cut and fixed in the 2% glutaraldehyde phosphate buffer (pH 7.2) for 30 min at room temperature, followed by washing and postfix in 1% *w/v* OsO4 with 0.05 M K3Fe(CN)6 in 0.12 M phosphate buffer (pH 7.2) for 2 h. After that, a standard procedure for dehydration, embedding, and sectioning was applied. Finally, the ultra-thin sectioned sample (~60 nm) were collected on copper grids with Formvar supporting membranes, stained with uranyl acetate and lead citrate, and examined by a Philips CM-100 electron microscope (Philips, Eindhoven, The Netherlands) operated at 100 kV. For SEM, specimens were sandwiched in 2 × 100 μm planchettes and cryopreserved by high-pressure freezing (HPM100, Leica, Vienna, Austria). The sandwiched planchettes were mounted in a planchette holder (Leica) under liquid N2 and transferred to a vitreous cryo transfer shuttle (VCT100, Leica). The samples were cracked and sputter-coated (approximately 6 nm) (MED020, Leica) with carbon/platinum. Specimens were examined with an FEI Quanta 3D scanning electron microscope operated at an accelerating 2 kV voltage.

#### *2.3. Preparation of Emulsions for In Vivo Studies*

From the initial five emulsifiers tested in the in vitro system, we selected WPC-PL and SL for in vivo study 1 and 2. While WPC-PL was chosen because it had shown a beneficial effect in infant formula [32], we chose SL as it is a common emulsifier and stabilizer often used in infant formulas. In the in vivo study, three were used as the most promising experimental emulsifier (WPC-A-EV) along with WPC-PL and SL to validate the improved digestibility observed in vitro. Ideally, all three emulsifiers could have been studied in each in vivo experiment but it was not feasible to include so many groups as the studies were very labor-intensive. To prepare the emulsions, we used a Rannie homogenizer (APV, Copenhagen, Denmark) at pressure 25 bar/250 bar, instead of a microfluidizer. The oil-in-water emulsions for the in vivo studies were made with 10% oil (*w*/*w*%) using an oilblend containing 91.66% Akonino NS, 5.46% MEG-3, and 2.88% Arasco, and 1% emulsifier. Different experimental diets were used for each of three in vivo studies: In in vivo study 1, the experimental diet consisted of complete formulas based on 10% oil-in-water emulsions

with either SL (VWR, Darmstadt, Germany) or WPC-PL mixed with a base-diet to achieve a final fat concentration of 5.1%. The base-diet was made from whey protein (Lacprodan® DI-9224), casein (Miprodan® 40), lactose, and minerals (Variolac® 855, all Arla Foods Ingredients) and was designed to meet the nutritional needs of pigs. In in vivo study 2, the experimental diet consisted of pure 10% oil-in-water emulsions with either SL (VWR) or WPC-PL. Finally, in in vivo study 3, the experimental diet consisted of complete formulas based on 10% oil-in-water emulsions with either SL (AAK, Aarhus, Denmark), WPC-PL, or WPC-A-EV mixed with the base diet to a final fat concentration of 10%.

#### *2.4. In Vivo Lipid Digestibility of Complete Formulas (Study 1)*

All procedures were approved by the Danish Animal Experiments Inspectorate (license number 2014-15-0201-00418), which follows the guidelines from Directive 2010/63/EU of the European Parliament and the ARRIVE guidelines [33]. We used cesarean-derived preterm neonatal piglets as they were assumed to have lower fat digestive capacity relative to term pigs, thereby making any effect of emulsifiers more detectable. In brief, one litter of preterm piglets, *n* = 22, (Landrace × Large white × Duroc, Gadstrup, Denmark) was born by cesarean section at day 113 of gestation and reared in preheated and oxygenated incubators as described previously [34]. Immediately after birth, the pigs were equipped with oral and vascular catheters to allow enteral and parenteral feeding. See supplemental for further information. On the second day, the pigs were surgically equipped with a jejunostomy to allow the quantitative collection of stoma output. Details for housing, feeding, surgery, post-surgical care, and sample collection are provided in the supplemental material. The piglets were block-randomized according to bodyweight into two groups receiving complete formulas emulsified with either SL (*n* = 5–9) or WPC-PL (*n* = 6–9). Enteral feeding was initiated as quickly as possible postoperatively at a rate of 6 mL/kg every three hours, gradually increasing to 15 mL/kg every three hours on day seven and eight. The personnel were blinded to the treatment groups. Stoma output was collected quantitatively on days 3, 4, and 7, and following measurement of fat concentration in the stoma output, intestinal fat absorption was calculated as described earlier [35].

Fat accumulation in the tissues of the small intestine was measured. A piece of the proximal part of the small intestine was fixed in a cryo cassette (Tissue-Tek Cryomold, Sakura Finetek, Zoeterwouder, Holland, The Netherlands) with tissue O.C.T. (Tissue-Tek, Sakura, Finetek Zoeterwoude, Holland) and frozen in liquid nitrogen. The tissue was sliced with a cryostat (Leica CM 1950, Leica Biosystems, Wetzlar Germany), and stained with Oil-Red-O (Sigma-Aldrich, Darmstadt, Germany). Six digital images were taken at 20× magnification from six different regions using a microscope (Olympus BX45, Tokyo, Japan), and the degree of fat infiltration was scored.

#### *2.5. In Vivo Fat Absorption Kinetics of Pure Emulsions (Study 2)*

Subsequently, we examined the triglyceride absorption kinetics after dosing of pure 10% oil-in-water emulsions. One litter of preterm piglets, *n* = 19, was born by cesarean section at day 106 of gestation, making the piglets 7 days more premature relative to the first study which further sensitizes the gut toward low digestive capacity. We have used this degree of prematurity in many previous experiments and based on sensitivity to develop prematurity-related complications like necrotizing enterocolitis, which may correspond to week 25–28 in human pregnancy. Procedures for cesarean section, postnatal rearing, and provision of parenteral nutrition were identical to the previous study. On day two, the piglets were blocked randomized according to body weight to receive an enteral bolus (6 mL/kg body weight) of pure emulsion with either SL or WPC-PL. Using a cross-over design, the piglets received a second pure emulsion bolus on day three, such that each piglet had been exposed to both emulsions. Blood samples from the arterial umbilical catheter were collected on both days at t = 0, 90, 180, 270, 360, and 540 min after the bolus was given and stored in heparinized tubes. Plasma was isolated after centrifugation (1270× *g*, 4 ◦C, 10 min), and the concentration of triglycerides was analyzed using an automated ADVIA 1800 Chemistry System (Siemens Healthcare A/S, Ballerup, Denmark). Following blood sampling, all pigs were euthanized.

#### *2.6. In Vivo Lipid Absorption Kinetics from Complete Formulas (Study 3)*

In a final in vivo experiment, the triglyceride absorption kinetics following ingestion of complete formulas was determined. A total of 49 preterm piglets from three pregnant sows were born by cesarean section at day 106 of gestation. They were stabilized and received parenteral nutrition as described above and were block-randomized according to birthweight into three groups, all receiving the complete formula, with a fat content of 10%. Collectively for all three in vivo studies, we ensured to have pigs from each litter equally represented in all treatment groups, allowing us to correct for any variance between litters and their specific genotype, in the analysis of variance. The complete formulas were made with the emulsions based on SL, WPC-PL, or WPC-A-EV. To stimulate the intestinal absorptive function, we initially fed the pure base-diet without emulsions during the first 24 h, at a rate of 3 mL/kg every three hours. On day two, the complete formulas including the emulsions were given at a bolus dose of 9 mL/kg, following a 6 hr fasting period. Blood samples were collected via the umbilical catheter at t = 0, 30, 60, 90 min after the bolus. Following this, the piglets were again fed with increasing amounts of enteral diet (6–9 mL/kg every three hours). On day four, the piglets were again fed a test meal of 9 mL/kg after a fasting period of 3.5 h, and blood samples were collected at t = 0, 30, 60, 90, and 120 min. Blood was drawn via the jugular vein in cases where the umbilical catheter was dysfunctional. Plasma was isolated, and triglyceride concentration was measured.

The pigs were euthanized on day four following a standardized feeding regimen to ensure an equal amount of gastric content at the time of euthanasia. Specifications for recordings and sample collection are provided in the supplementary section. Gastric content was weighed, and gastric lipase activity in gastric content was measured using the method described earlier with slight modifications [36]. The assay was carried out using the same pH Stat equipment as for the in vitro lipolysis. It was initiated by mixing 14.5 mL assay medium containing 1.5 μM Bovine Serum Albumin (AppliChem, Darmstadt, Germany), 150 mM NaCl (VWR), and 2 mM sodium taurodeoxycholate (Sigma Aldrich) with 0.5 mL tributyrin (Sigma Aldrich) and 0.5 mL stomach content. The enzymatic activity, calculated as U/mL was based on the measured titration rate of NaOH by a Metrohm Titrando pH Stat over five minutes of digestion at pH 5.5. As the butyric acid was not being fully titratable at pH 5.5, a correction factor of 1.12 was multiplied to the calculated activity. The total amount of collected stomach content, as well as the pH, was also measured.

#### **3. Statistics**

Statistical analyses for the in vitro digestions were performed using GraphPad Prism version 7.0 for Windows (GraphPad Software, San Diego, CA, USA, www.graphpad.com). Group comparisons were made using unpaired t-tests based on the area under the curve (AUC).

Statistical analyses for the in vivo studies were performed using R (version 3.5.0, R Foundation for Statistical Computing, Vienna, Austria). Continuous data were analyzed with a linear mixed model using the lm function (lme4 package). Group comparisons were made with an ANOVA (lme4 package) and Post Hoc Tukey test with the glht function (multcomp package). Normal distribution and homoscedasticity of residuals were visualized for model validation. Birthweight was used as a covariate, sex as a fixed variable, and litter as a random variable in all the models. Repeated blood samples were analyzed using SAS (SAS Software 9.4, SAS Institute, Cary, NC, USA). In all repeated measures models, birthweight and baseline triglyceride levels were included as covariates, whereas sex and litter were included as fixed and random effects respectively. Survival curves of basic motor skills were evaluated in GraphPad (Prism version 7.0) using the Logrank test for the trend. Fat accumulation scores were analyzed as ordinal data with the clmm function (ordinal package), and group comparison was made with the post hoc Tukey test. Group comparisons of the gastric lipase activity were made with unpaired t-tests in the GraphPad (Prism version 7.0). *p*-values < 0.05 were regarded significant and *p*-values < 0.1 as a tendency to effect. Data are presented as means ± standard deviation unless otherwise stated.

#### **4. Results**

#### *4.1. In Vitro Lipolysis*

The in vitro lipolysis assessed as the amount of released free fatty acids (FFAs) titrated over time during digestion of each of the five 3.5% fat emulsions, using SL, WPC-PL, WPC-A-EV, WPC-A-TAG, and WPC-A-WP, is presented in Figure 1. After 50 min of in vitro digestion 28–44 μmol of FFAs were titrated depending on the emulsifier (SL: 28.1 μmol ± 0.5, WPC-A-WP: 29.3 μmol ± 1.6, WPC-A-TAG: 38.0 μmol ± 1.7, WPC-PL: 42.2 μmol ± 0.6, and WPC-A-EV: 44.3 μmol ± 6.6). After 140 min 145–190 μmol of FFAs were titrated (SL: 145.2 μmol ± 6.4, WPC-A-TAG: 148.6 μmol ± 13.7, WPC-PL: 176.8 μmol ± 1.1, WPC-A-WP: 180.6 μmol ± 8.0, and WPC-A-EV: 190.2 μmol ± 2.0). The in vitro digestion of emulsions based on WPC-A-EV, WPC-PL, and WPC-A-WP showed a higher lipolysis rate relative to SL. This was found, both when considering the AUC of the gastric and intestinal step separately, as well as the two combined (all *p* < 0.01), whereas WPC-A-TAG showed values similar to SL.

**Figure 1.** Titration of free fatty acids during simulated human gastric digestion with recombinant human gastric lipase (rHGL) and pepsin at pH 6.4 (0–50 min) and pH 6.5 (50–140 min) of 3.5% oilin-water emulsions using either whey protein concentrate from acid whey-enriched in extracellular vesicles (WPC-A-EV) (*n* = 3), WPC enriched in phospholipids (WPC-PL) (*n* = 3), whey protein concentrate from acid whey enriched in triglycerides (WPC-A-TAG) (*n* = 3), whey protein concentrate from acid whey enriched in soluble whey protein (WPC-A-WP) (*n* = 3), and soy lecithin (SL) (*n* = 2 due to removal of outliers). Values are presented as mean ± SD. \*\* *p* < 0.001. \*\*\* *p* < 0.0001.

#### *4.2. In Vitro Study 2—Microstructure of Emulsions*

The droplet size distribution for each of the five 3.5% fat emulsions, SL, WPC-PL, WPC-A-EV, WPC-A-TAG, and WPC-A-WP, are presented in Figure 2, Panel 1 as the volume mean diameter. All emulsions had droplet sizes within the range of 0.04–100 μm, with the smallest droplet size being observed for the SL emulsion, which had a unimodal distribution centered at 0.16 μm. WPC-A-EV and WPC-PL emulsions had similar droplet size distributions, and both had a primary population with modes centered at 0.5–0.7 μm and a smaller population around 7 μm.

The droplet sizes in the WPC-A-TAG and WPC-A-WP emulsions were similar and showed two distributions of nearly uniform height, with the first at 0.6 μm for both WPC-A-TAG and WPC-A-WP and the second mode around 4 μm for WPC-A-TAG and 17 μm for WPC-A-WP. The microstructure of WPC-PL, WPC-A-EV, and SL emulsion was further characterized by TEM (Figure 2, Panel 2). In general, the lipids droplets in the WPC-PL (subpanels A and D) and WPC-A-EV (subpanels B and E) emulsions had a similar structure with clear edges and relatively round and smooth appearances compared with droplets in the SL emulsion (subpanels C and F). Some of the droplets in the WPC-PL emulsions were partially covered by a thick dark layer (subpanel A), which might be composed of milk protein (mainly from aggregated whey proteins) from the WPC-PL product. These thick

layers were less frequently observed for the droplet made of WPC-A-EV (subpanel B). Due to the higher content of native whey proteins present in WPC-A-EV relative to WPC-PL (communication with Arla Foods Ingredients), less aggregated protein complex can be associated with the interface of the lipid droplet presumably due to lower protein content in WPC-A-EV relative to WPC-PL. The SL lipid droplets had a different morphology than that of the other two emulsions, with irregular and uneven appearances (subpanel C). Besides the fluffy and thick layer of phospholipids on the droplet's surface, we also observed a white layer of particles that either lies on the surface of the lipids droplets or are incorporated within the droplets. These droplets may be nano-sized liposomes generated during homogenization. Further, the cryo-SEM images provided a 3D cross-section view of the emulsions, which indicated that the surface of lipid droplets in the WPC-PL and WPC-A-EV emulsions were thinner and smoother relative to the SL emulsion droplets.

**Figure 2.** Particle size distribution and transmission electron microscopy (**Panel 1**): Volume mean diameter for each of the 3.5% oil-in-water emulsions, WPC-A-EV, WPC-PL, WPC-A-TAG, WPC-A-WP, and soy lecithin (SL). *n* = 3. Values are presented as means ± SD. (**Panel 2**): Micrographs from transmission electron microscopy (subpanels A, B, C) and cryo-scanning electron microscopy (subpanels D, E, F) of three emulsions. A and D: WPC-PL emulsion, B and E: WPC-A-EV emulsion, C and F: SL emulsion. The scale bar for micrographs A–C is 500 nm, and D–F is 2 μm. The arrow in A points out the dark layer, in E the arrow points out the lipid droplet's smooth surface, and in F the arrow points out the thick and uneven surface of the lipid droplet.

#### *4.3. In Vivo Study 1—In Vivo Fat Digestibility of Complete Emulsions*

Of the initial 22 piglets, three were euthanized due to post-surgical complications, and later two died from poor clinical conditions. The final number of piglets included in each group was *n* = 9 (SL) and *n* = 8 (WPC-PL). Birthweight was similar between WPC-PL and SL piglets, with a mean weight of 1530 g ± 235 and 1465 g ± 306, respectively. Changes in body weight in the immediate postnatal period and following period with an ileostomy were generally characterized by weight loss in both groups, albeit with lower weight loss in WPC-PL versus SL (−12.4 ± 12.3 versus −17.5 ± 14.9 g/(kg·day), *p* > 0.05). We were able to quantify stoma output reliably in a subfraction of the piglets (*n* = 5–10 piglets),

see Figure 3. Based on this, we found that the fat absorption coefficient was similar for WPC-PL and SL (95.4% ± 2.5 in WPC-PL versus 95.8% ± 2.3 in SL) on day three, declining to 85.7% ± 9.4 in WPC-PL versus 83.2% ± 14.26 in SL on day seven.

Due to poor clinical conditions, three piglets did not receive their last feeding, and they were excluded. Plasma triglyceride levels one hour postprandial did not differ between groups, reaching a level in the WPC-PL piglets (*n* = 7) of 0.25 mmol/L ± 0.12 and in the SL piglets (*n* = 6) of 0.15 mmol/L ± 0.03. Organ weights and histological fat accumulation score in the proximal intestine were similar between the groups. Details are provided in Supplementary Table S1 and Figure S1.

#### *4.4. In Vivo Study 2—In Vivo Fat Absorption Kinetics of Pure Emulsions*

Using the cross-over design, the final group sizes were *n* = 16 for WPC-PL and *n* = 18 for SL. Mean body weight at birth was 933 g ± 277 (WPC-PL) and 969 g ± 226 (SL). Triglyceride levels in plasma were similar for WPC-PL and SL on both days two and three, and for the pooled values across both days (Figure 4).

Peak plasma triglyceride level was reached between 60–120 min after bolus, with no noticeable difference between the groups. Other plasma measurements, including cholesterol, glucose, ALAT, ASAT, and creatinine, were also similar for WPC-PL and SL. The only exception from this was the WPC-PL piglets' baseline level of total bilirubin, which was significantly higher relative to SL, *p* = 0.03. All biochemistry data are available in Supplementary Table S2.

**Figure 4.** Triglyceride levels in plasma, mmol/L, as pooled data across day two and three in upper panel (WPC-PL *n* = 16, SL *n* = 18), and data from day two in middle panel (WPC-PL *n* = 9, SL = 9) and three in lower panel (WPC-PL *n* = 9, SL *n* = 7) separately. Values presented as means ± SD.

*4.5. In Vivo Study 3—In Vivo Fat Absorption Kinetics from Complete Formulas*

Final number of piglets were *n* = 17 (SL), *n* = 14 (WPC-PL) and *n* = 15 (WPC-A-EV). Birthweight was similar between SL (1124 g ± 264), WPC-PL (1077 g ± 225) and WPC-A-EV (1108 g ± 260) fed piglets, weight gain was similar for SL (−5.39 g/(kg·day) ± 11.6), WPC-PL (−2.19 g/(kg·day) ± 10.1) and WPC-A-EV (−1.24 g/(kg·day) ± 7.4). Three piglets were euthanized due to clinical complications unrelated to the diets.

Following the bolus administration on day two, plasma triglyceride levels increased similarly for all three groups over the entire 90 min sampling time relative to the common baseline level. (Figure 5, upper panel). When the bolus test was repeated on day four, plasma triglyceride level again peaked between 60–90 min, and importantly, the WPC-A-EV and WPC-PL groups had significantly higher plasma triglyceride levels compared to the SL group, *p* < 0.01 and *p* < 0.001 respectively. The values for WPC-A-EV and WPC-PL were similar (Figure 5, lower panel).

#### *4.6. Gastric Fat Concentration and Lipase Activity—Related to In Vivo Study 3*

As a result of the standardized feeding regimen before euthanasia, the gastric residuals were very similar across the groups, i.e., 15–16 g per pig (Supplementary Table S3). However, the fat concentration in the residuals was significantly higher in the SL piglets: 157 ± 27 mg/g, compared with WPC-PL: 108 ± 17 mg/g and WPC-A-EV: 103 ± 13 mg/g, *p* < 0.001 in both cases. Due to higher fat content in the SL group gastric residuals, the subsequent analysis of gastric lipase showed the highest activity in the SL group relative to WPC-PL and SL, which may partly be due to the higher availability of substrate (i.e., fat) for the lipase assay. Specific numbers were SL: 0.92 U/mL ± 0.63; WPC-PL: 0.27 U/mL ± 0.15 and WPC-A-EV: 0.30 U/mL ± 0.34 (*p* < 0.01 for WPC-PL and WPC-A-EV relative to SL). The gastric pH was similar between groups (Figure 6).

**Figure 5.** Triglyceride levels in plasma at day two (WPC-A-EV *n* = 16, WPC-PL *n* = 16, SL *n* = 17) and four (WPC-A-EV *n* = 15, WPC-PL *n* = 14, SL *n* = 17), baseline sample and 30, 60, 90 and 120 min after a test bolus was given. Presented as mean ± SD. \* *p* < 0.01. \*\* *p* < 0.001.

**Figure 6.** (**Top panel**): Gastric pH for WPC-A-EV (*n* = 14), WPC-PL (*n* = 11), and SL (*n* = 17) in gastric content collected post euthanasia. (**Bottom panel**): Gastric lipase activity for WPC-A-EV (*n* = 13), WPC-PL (*n* = 11), and SL (*n* = 16) in gastric content collected post euthanasia. \*\* *p* < 0.01.

Further, circulatory levels of albumin, liver enzymes, creatinine, creatine kinase, glucose, phosphate, and urea, were largely similar across the groups (Supplementary Table S4).

#### **5. Discussion**

The most important finding was that piglets fed complete formula with WPC-PL and WPC-A-EV emulsifiers showed higher plasma triglyceride levels relative to the SL group when studying fat absorption kinetics in vivo. This notion was further substantiated in the in vitro lipolysis assays. MFGM surrounds the milk fat globule in human milk, and this membrane structure has its unique content of phospholipids, glycerolipids, cholesterol, and glycosylated membrane proteins. This provides physical properties allowing lipases to have high catalytic efficiency during gut luminal fat digestion [26]. In contrast, the fat fraction in infant formula is emulsified primarily by SL and milk proteins like β-lactoglobulin and caseins, which creates a different lipid-water interface compared with MFGM in human milk [29]. Therefore, we hypothesized that the utility of naturally occurring dairy emulsifiers isolated industrially from bovine whey, here referred to as WPC-PL, WPC-A-TAG, WPC-A-WP, and WPC-A-EV, has superior absorptive effects relative to SL in vitro and in vivo. These absorptive effects could potentially result in higher availability of lipids for brain development. Gastric lipase activity is particularly important in early life, and surprisingly our ex vivo analysis of gastric lipase activity in the stomach content showed the highest activity in the SL group relative to WPC-PL and WPC-A-EV. This may, however, partly be explained by a higher residue of undigested lipids in the SL group gastric residuals, leading to more substrate available for the ex vivo lipolysis assay. However, the gastric fat content was approximately 50% higher in the SL group relative to WPC-PL and WPC-A-EV. We normalized the gastric lipase activity to gastric fat content and found that it does not fully account for the higher lipase activity seen in the SL group, and we speculate that other regulatory mechanisms play a role.

An adequate supply of lipids, including LC-PUFA to the brain during early life, helps secure synaptogenesis and brain maturation and gives lasting effects on cognition [21]. Therefore, infant formulas should ensure LC-PUFA is available to the brain. Although efforts have been made to customize infant formulas to mimic human milk, including fortifying formula with LC-PUFA, the results have been inconclusive with regard to cognitive effects [20]. Fat absorption is lower in formula-fed newborn infants than babies fed mother's milk, and assuming that the level of endogenous digestive enzyme activity is similar between breastfed and formula-fed infants, this indicates that factors that relate to the fat fraction per se, determine the level of lipolysis and absorption.

The interfacial layer of the lipid droplets affects the fat digestion in milk, and our results indicate that when the interfacial layer is composed mainly of phosphatidylcholine (i.e., the SL emulsions), the lipolysis is slower compared to lipid droplets emulsified with bovine dairy emulsifiers of more complex polar lipid and protein composition. This agrees with the findings from earlier studies suggesting that the pancreatic lipase has a higher affinity for a lipid surface covered with casein or whey proteins [37]. This notion is supported by the TEM and cryo-SEM microstructure micrographs, which indicate that the surface of WPC-PL and WPC-A-EV emulsified droplets is thin and smooth, which may favor the lipase to penetrate and reach the lipids. In contrast, in SL emulsions, the formation of liposomes may additionally reduce the lipase efficacy.

We used a neonatal preterm piglet model to determine the effects of WPC-PL, WPC-A-EV, and SL emulsifiers on the absorption of triglycerides. Appropriate animal models offer the potential of eliminating important confounders that exist in the comparison of breastfeeding versus formula feeding, and we suggest that preterm pigs with their resemblance to preterm infants in terms of gut and brain development [34] and post-surgical responses [38], is a valuable tool to determine these effects. Our initial study of triglyceride digestibility using an ileostomy model showed similar triglyceride absorption levels between the WPC-PL and SL groups. Plasma triglyceride levels measured one hour postprandial were numerically higher in the WPC-PL group, but this did not reach significance with the limited sample size. The

preterm pigs were born on gestational day 113, i.e., 4–5 days before the expected termination date. It is possible that the degree of prematurity was too small to sensitize the digestive system toward reduced-fat uptake. In the two subsequent piglet studies, we chose to further sensitize the model by delivering the piglets on day 106 of gestation, i.e., 11–12 days before the expected term date. Yet, histological lipid stainings showed only numerically higher scores for WPC-PL and WPC-A-EV relative to SL. From the notion that fat uptake is indeed higher in WPC-PL and WPC-A-EV relative to SL, the apparently similar fat level in the mucosa may indicate that the transport of fat as chylomicrons into lymph vessels is an equally efficient process relative to absorption from the gut lumen, resulting in no net fat accumulation in the mucosa.

Several randomized controlled trials have examined the safety and efficacy of dairy phospholipid-enriched infant formulas. In one multicenter randomized controlled trial, they used two commercial phospholipid-rich ingredients and investigated its safety in infant formula [39]. It was proven safe to use in healthy term-born infants, with eczema being the only adverse effect reported. Likewise, Xionan and colleagues [40], found that a commercial dairy phospholipid-rich ingredient was safe and well-tolerated, and found no difference in skin effects relative to neither standard formula nor breastfeeding. In another randomized controlled trial, a standard formula was compared with an experimental formula enriched with the same dairy phospholipid-rich ingredient and showed improved cognition score in the experimental group using Bayley Scales of Infant and Toddler Development III at 12 months of age [32] and lower incidence of acute otitis media [41]. Whereas the dairy phospholipid-enriched group generally performed similar to a breastfed reference group, the breastfed infants performed better than both formula-fed groups on a verbal subscale. In the same infants, an analysis of lipidomics profile in serum/plasma and erythrocyte membranes at four, six, and twelve months of age showed significant differences between the dairy phospholipid-enriched formula and the standard formula groups. The difference in serum did, however, disappear six months after the intervention [42]. The discrepancy in the lipidomics profile was mainly accounted for by sphingomyelin, likely to be explained by the contribution from the dairy phospholipid ingredient. The provision of sphingomyelin into the blood circulation may provide the developing brain with this essential nutrient, yet the direct causality remains to be determined. In support of the notion that dairy phospholipids are important for brain development, Gurnida and colleagues showed positive effects of glycosphingolipids on hand-eye coordination, performance score, and total score measured with the Griffiths Mental Developmental Scale (GMDS) in term-born infants [43].

We conclude that the milk-derived emulsifiers WPC-PL and WPC-A-EV increase fat digestion and absorption of triglycerides relative to SL. The effects are seen in vivo when the emulsions are an integrated part of a complete diet, but not when the pigs were fed pure emulsions. Higher levels of lipid hydrolysis, as indicated in vitro, suggest that the higher absorptive rates in vivo result from increased lipase activity when WPC-PL and WPC-A-EV are used as emulsifiers relative to SL. It remains to be determined whether emulsification with these milk-derived polar lipids can affect brain development.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-664 3/13/2/410/s1, Table S1. Organ weight of pigs equipped with an ileostomy and fed milk diets with SL or WPC-PL for seven days; Table S2. Circulatory markers in pigs with boluses of pure emulsion; Table S3. Organ weight in pigs fed complete formulas with either SL, WPC-PL, or WPC-A-EV; Table S4. Circulatory markers in pigs fed complete formulas with either SL, WPC-PL, or WPC-A-EV as an emulsifier; Figure S1. Higher score indicates higher fat infiltration, as assessed subjectively by several independent observers who were blinded to the treatment groups [44].

**Author Contributions:** Conceptualization: K.B.K.K., C.H., P.N., A.M., P.T.S., M.S.O. and T.T.; Methodology: K.B.K.K., C.H., T.R.S.J., X.G., P.B.J., I.I. and T.T. Resources: N.D., P.N. and M.S.O. Data analysis: K.B.K.K., C.H., T.R.S.J., X.G. and T.T.; Writing original draft: K.B.K.K., C.H., T.R.S.J. and T.T.; Writing, review and editing: all. Project administration: University of Copenhagen; Funding aqusition: Denmark; T.T. had primary responsibility for final content. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by The Innovation Foundation Denmark (Grant No.: 5158- 00014B), the Danish Dairy Research Foundation and Arla Foods Ingredients.

**Institutional Review Board Statement:** Danish Animal Experiments Inspectorate (license number 2014-15-0201-00418).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All generated data are stored on pass word protected servers at University of Copenhagen, and can be made available upon request.

**Acknowledgments:** We thank Elin Skytte, Jane Povlsen, Kristina Møller, Michelle Christiansen, Rikke Stolberg and Anne Heckmann for their valuable contributions. The late Lars Ingvar Hellgren, Technical University of Denmark, is acknowledged for his pas-sionate initiation and steering of the Infant Brain project until his health condition prevented further involvement.

**Conflicts of Interest:** Authors M.S.O., X.G., N.D. are employees of Arla Foods Ingredients. Remaining authors have no conflicts of interest.

#### **References**


## *Article* **Acute Effects of Cheddar Cheese Consumption on Circulating Amino Acids and Human Skeletal Muscle**

**Naomi M.M.P. de Hart 1,†, Ziad S. Mahmassani 2,†, Paul T. Reidy 3, Joshua J. Kelley 2, Alec I. McKenzie 4, Jonathan J. Petrocelli 2, Michael J. Bridge 5, Lisa M. Baird 6, Eric D. Bastian 7, Loren S. Ward 8, Michael T. Howard <sup>6</sup> and Micah J. Drummond 1,2,\***


**Abstract:** Cheddar cheese is a protein-dense whole food and high in leucine content. However, no information is known about the acute blood amino acid kinetics and protein anabolic effects in skeletal muscle in healthy adults. Therefore, we conducted a crossover study in which men and women (*n* = 24; ~27 years, ~23 kg/m2) consumed cheese (20 g protein) or an isonitrogenous amount of milk. Blood and skeletal muscle biopsies were taken before and during the post absorptive period following ingestion. We evaluated circulating essential and non-essential amino acids, insulin, and free fatty acids and examined skeletal muscle anabolism by mTORC1 cellular localization, intracellular signaling, and ribosomal profiling. We found that cheese ingestion had a slower yet more sustained branched-chain amino acid circulation appearance over the postprandial period peaking at ~120 min. Cheese also modestly stimulated mTORC1 signaling and increased membrane localization. Using ribosomal profiling we found that, though both milk and cheese stimulated a muscle anabolic program associated with mTORC1 signaling that was more evident with milk, mTORC1 signaling persisted with cheese while also inducing a lower insulinogenic response. We conclude that Cheddar cheese induced a sustained blood amino acid and moderate muscle mTORC1 response yet had a lower glycemic profile compared to milk.

**Keywords:** dairy; ribo-seq; muscle protein synthesis; anabolism; insulin

#### **1. Introduction**

Aminoacidemia from the digestion of protein sources is a major stimulator of skeletal muscle protein anabolism and important for maintenance of muscle mass and overall muscle health. Circulating amino acid kinetics and acute skeletal muscle protein anabolic responses have been extensively evaluated following ingestion of dairy proteins such as casein and whey protein isolate [1–5]. Though these data have provided fundamental information in understanding how muscle responds to protein, it is less generalizable to

**Citation:** de Hart, N.M.M.P.;

Mahmassani, Z.S.; Reidy, P.T.; Kelley, J.J.; McKenzie, A.I.; Petrocelli, J.J.; Bridge, M.J.; Baird, L.M.; Bastian, E.D.; Ward, L.S.; et al. Acute Effects of Cheddar Cheese Consumption on Circulating Amino Acids and Human Skeletal Muscle. *Nutrients* **2021**, *13*, 614. https://doi.org/10.3390/ nu13020614

Academic Editor: Dennis Savaiano

Received: 15 January 2021 Accepted: 10 February 2021 Published: 13 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

the community since most dietary protein sources contain a mixed-macronutrient profile, contain many micronutrients within their matrix, and are more complex during digestion.

Recent protein metabolism studies have evaluated blood amino acid kinetics and muscle anabolic responses to protein-enriched, nutrient-complex foods such as beef, egg, and pork [6–14] and as a result, have demonstrated unique amino acid and protein anabolic responses. For example, consumption of 18 g of protein from whole egg after a bout of exercise increased protein synthesis more so than egg whites in spite of similar post absorptive plasma leucine levels [14]. This suggests protein-dense whole foods have utility to promote protein anabolism not simply predicted by the amount of protein or level of aminoacidemia, which is in contrast to what has been observed with isolated protein products [5]. Therefore, there is a continued need to characterize whole food products to identify high quality protein sources that encourage human health.

To our knowledge, no studies have evaluated the amino acid pattern in plasma or muscle anabolic response to cheese ingestion. Cheddar cheese, is a low carbohydrate, high-fat, protein-rich food that is a regular dietary component of the U.S. diet [15]. Cheddar cheese has a well characterized amino acid profile with a high content of leucine (~10%) and is considered low glycemic. Moreover, the protein in Cheddar cheese is partially hydrolyzed due to aging/ripening [16], and therefore is likely to speed up digestion and promote the appearance of amino acids in the circulation [4,17]. Cheddar cheese is also composed of many other underappreciated nutrients within its food matrix [18] that are beneficial for human health and could further enhance protein anabolism.

Therefore, the primary purpose of this study was to characterize the amino acid response following 65 g (20 g protein) of Cheddar cheese, an amount of protein capable of increasing blood amino acid levels from a whole dairy product [7,9,19]. In addition, to gain insight on the protein anabolic processes in skeletal muscle, we evaluated mTORC1 localization and cellular signaling following cheese ingestion, given that mTORC1 intracellular signaling is highly responsive to acute protein intake particularly to sources that are rich in leucine [20,21]. We also complimented mTORC1 signaling with a unique 'omics approach of ribosome profiling [22] to capture key information regarding which mRNAs are translated after cheese ingestion. Finally, to provide context in comparison to a well-described whole food, we conducted a within subject crossover study comparing to an isonitrogenous amount of milk [19]. We hypothesized that a single dose of Cheddar cheese in young male and female adults, equivalent to 20 g of protein, would acutely increase the blood branched-chain amino acids (particularly leucine) and induce a translational program characterized by mTORC1 signaling.

#### **2. Methods**

#### *2.1. Subjects*

Twenty-four young male (*n* = 12) and female (*n* = 12) subjects participated in this study (Table 1; 27 ± 4 years; BMI 23.1 ± 3.5 kg/m2). Interested subjects were notified of the study through posted flyers on campus and in areas around the university and were also contacted through the University of Utah PEAK Health and Fitness registry. Subjects were screened (self-report) based on the following exclusion criteria: history of cardiovascular disease, endocrine or metabolic disease (e.g., hypo/hyperthyroidism, diabetes), kidney disease or failure, liver disease, respiratory disease (acute upper respiratory infection, chronic lung disease), stroke with motor disability, use of anticoagulant therapy (e.g., Coumadin, heparin) including aspirin and fish oils within 7 days (d) of the first metabolic experiment, elevated systolic blood pressure > 150 or a diastolic blood pressure > 100, smoking, recent anabolic or corticosteroids use (within 12 weeks of first biopsy), pregnancy, lactose intolerance, and irregular menstruation. Enrolled participants read and signed the informed consent document, which was approved by the University of Utah Institutional Review Board (IRB #110963) and in agreement with the Declaration of Helsinki. This study is registered at clinicaltrials.gov (NCT04660877).



Mean ± SD, \* Different from Female (*p* < 0.05).

#### *2.2. Experimental Design*

After enrollment, participants completed baseline testing which included a dietary assessment, body composition and habitual activity levels. Body composition (lean and fat mass) was assessed using a Bod Pod instrument (conducted prior to Metabolic Study #1). Physical activity was tracked for a 7 days period between the Metabolic Study visits. Additionally, a 3 d daily dietary record (ASA24) was recorded before each Metabolic Study visit. The daily dietary record was averaged between all recorded days and reported in Table 1.

Each subject took part in two metabolic studies (Figure 1) with each designed to test the acute blood and muscle response to an ingested amount of either Cheddar cheese or milk matched for protein (Table 2). Approximately, one month after the first experiment (Metabolic Study #1), the participant completed the second experiment (Metabolic Study #2) which was exact in design and at the same time of day as the first study but the participant ingested the alternate food product. Prior to each of the metabolic studies, the participant ate a standardized research meal the night before the study and refrained from intense physical activity for 48 h. The morning of the metabolic studies, the participant arrived at the clinical research center after a ~10 h fast. A catheter was then placed in the participants' arm for blood sampling. Next, the participant underwent a baseline vastus lateralis skeletal muscle biopsy (0 min) using a modified version Bergström muscle biopsy technique [23]. Following the baseline muscle biopsy, the participant consumed either Cheddar cheese (65 g) or milk (370 mL; 2%; Fairlife) each amounting to 20 g of protein. The Cheddar cheese was processed at Glanbia Nutritionals, aged to one month, and frozen into batches distributed monthly by the sponsor as needed. The amino acid profile of low-fat Cheddar cheese and 2% Fairlife milk can be found in Supplemental Table S1. Subsequent muscle biopsies occurred 60 and 180 min on the same thigh after product ingestion which is an ideal timeframe to capture mTORC1 signaling and mRNA translational events following proteinenriched nutrient ingestion [24,25]. Blood sampling occurred before and periodically after ingestion of the products (up to 300 min). Blood samples were taken every 20 min during the first 3 h and then every 30 min for the last 2 h. Therefore, there were a total of 14 blood draws and 3 muscle biopsies for each Metabolic Study visit. The starting thigh for muscle biopsies for the first Metabolic Study was randomized for each subject and balanced with the second Metabolic Study (left leg then right or right leg then left). Muscle samples were frozen in liquid nitrogen (for immunoblotting and ribosomal profiling) or prepared in O.C.T. (Optimal Cutting Temperature) and frozen in liquid nitrogen-cooled isopentane for the immunohistochemical assessment.

**Figure 1.** Overview of the crossover study experimental design.

**Table 2.** Nutrient content of experimental products.


#### *2.3. Blood Analyses*

Blood samples were collected in EDTA (Ethylenediaminetetraacetic) vacutainer collection tubes and immediately placed on ice. Samples were centrifuged (2500 rpm, 10 min) and plasma was collected and frozen at −80 ◦C until later analysis. Plasma was processed for essential and non-essential amino acids using the EZ:Faast Amino Acid Kit (Phenomenex; Cat #KG0-7165) and analyzed using GCMS analysis in collaboration with the institution's Metabolomics Core. Essential amino acids included detection of leucine, isoleucine, valine, threonine, methionine, phenylalanine, lysine, histidine, and tryptophan. Non-essential amino acids included detection of alanine, glycine, serine, proline, asparagine, glutamate, glutamine, and tyrosine. Samples were also immediately assessed for glucose (YSI) at the time of the study and later assessed for insulin (Human Insulin ELISA, Millipore Sigma, Burlington, MA, USA; EZHI-14K) and non-esterified fatty acids (NEFA-HR; Wako Chemicals, Richmond, VA, USA) in replicate using commercially available kits. Insulin and free fatty acids were determined at select time points (baseline, 20, 40, 80, 140, 210, and 300 min).

#### *2.4. Skeletal Muscle Immunoblotting*

Approximately 30 mg of tissue at each biopsy time point for Cheddar cheese and milk interventions was homogenized 1:10 (wt/vol) using a glass tube and mechanically-driven pestle grinder in an ice-cold buffer containing 50 mM Tris (pH 7.5), 250 mM mannitol, 40 mM NaF, 5 mM pyrophosphate, 1 mM EDTA, 1 mM EGTA, and 1% Triton X-100 with a protease inhibitor cocktail. Homogenates were centrifuged for 10 min at 4 ◦C. After centrifugation, the supernatant was collected and the protein concentration was determined using a modified Bradford protein assay and measured by a spectrophotometer (EPOCH; BioTek, Winooski, VT, USA).

Thirty micrograms of protein from muscle homogenate was separated via polyacrylamide gel electrophoresis, transferred onto a polyvinylidene difluoride membrane (PVDF), and incubated with primary and secondary antibodies. PVDF Membranes were imaged on a ChemiDoc XRS (Bio-Rad, Hercules, CA, USA) and quantified with Image lab software (Bio-Rad). The primary antibodies were purchased from Cell Signaling Technology and were the following: phospho-S6K1, Thr389, 1:1000, #9205; phospho-ribosomal protein S6, RPS6, Ser240/244, 1:1000, #2215; phospho-AS160, Ser588, 1:1000, #8730; phospho-GSK-3β, Ser9, 1:1000, #9336; phospho-Akt, Ser473, 1:1000, #9271. Secondary antibody (HRP Anti-Rabbit, #SC2004, 1:2000) was purchased from Santa Cruz Biotechnology. Phosphorylation of these proteins were normalized to Ponceau-S staining and reported as fold change from baseline.

#### *2.5. Skeletal Muscle Immunohistochemistry*

Muscle was sectioned into 8 μm cross-sections, mounted on slides in −25 ◦C, then left to air-dry overnight, and stored at −20 ◦C. Immunofluorescent staining was used to detect mTORC1 (Cell Signaling Technology, #2983, 1:100), the lysosomes (LAMP2: Abcam, #ab25631, 1:100), and the membrane (WGA: Fisher Scientific, #W32466, 1:50) as demonstrated by others [26–29]. Briefly, tissue was fixed in acetone (10 min), and the following blocking steps were performed: (1) endogenous peroxidases: 3% H2O2 for 7 min, (2) Non-Specific Binding Sites: 5% goat serum, Vector Labs #S-1000 with 0.3% Triton-X for 1 h, and (3) Avidin/Biotin: Vector Labs #SP-2001 according to manufacturer's instructions. WGA was added (5 min), and mTOR and LAMP2 were incubated on the slide overnight. Secondary antibody for LAMP2 was performed using Alexa Fluor 488 Tyramide SuperBoost (Invitrogen, #B40932, according to manufacturer's instructions), while secondary for mTOR was on Cy3 (Jackson ImmunoResearch, #711-165-152; 1:500) for 1 h. Finally, slides were mounted, cover slipped (Vectashield with DAPI, Vector Labs, #H-1200), and stored in the fridge until imaged (within 1 month of staining).

Images were taken using a Leica SP8 White Light laser confocal microscope equipped with automated stage, and Nikon NIS-Elements multi-platform acquisition software. At least 9 images (16 bit) were taken at 40X/1.3 magnification with oil immersion, with each image capturing ~5 muscle fibers per image in high detail at each time point, analyzing a total of ~45 muscle fibers per subject per time point, for each product consumed. When looking at events detected above threshold (set with help of combinations of positive and negative controls) of mTOR and LAMP2, anything not within 80% of the average was not used. The number of objects/events per channel times the average area covered by each object gave us the total area per channel. As previously described [26], Mander's overlap coefficient of colocalization was employed (k1 for mTOR/LAMP2; k2 for mTOR/WGA) to quantify the cellular overlap of these proteins, and this was performed in NIS-Elements for mTOR/LAMP2 and mTOR/WGA.

#### *2.6. Ribosomal Profiling*

Muscle samples at each time point (0, 60, 180 min) from Cheddar cheese and milk studies were assessed from a subset of subjects (4 subjects, 2 M, 2 F). Traditional RNA-Seq captures total mRNA abundance within a tissue sample, while the emerging technique of Ribo-Seq allows the capture of ribosome protected fragments (RPF) measuring translational

activity in a transcript-specific manner [30,31]. Polysome complexes were isolated, and unprotected mRNA digested with RNase I, and the ribosome protected mRNA footprints were analyzed by RNA-Sequencing methods as previously described by our group [22] with the exception that rRNA was removed from the RPF samples using the NEBNext rRNA Depletion kit and libraries were size selected by polyacrylamide gel electrophoresis on 6% native gels. Libraries were then sequenced on an Illumina Novaseq 6000 instrument. Raw sequence data can be obtained from the National Center for Biotechnology Information Gene Expression Omnibus repository entry GSE163279.

Uniquely mapping sequences were identified by alignments using bowtie to Reference Sequence database (RefSeq) mRNA entries obtained from the University of California, Santa Cruz browser (Hg38 human genome reference assembly) in which all mRNAs derived from the same gene were reduced to a single entry corresponding to the longest isoform. Normalization factors based on the trimmed mean of M-values were determined by using the calcNormFactors function of the Bioconductor package edgeR [32]. Dispersion estimates were obtained prior to likelihood ratio tests (glmFit and glmLRT functions of edgeR) to determine significance of the log2 fold change in RPFs or RNA for all transcripts with ≥1 count/million in all samples. Differences were considered significant if the false discovery rate was ≤0.05. Pearson's product-moment correlation coefficients were calculated.

Ingenuity Pathway Analysis was performed to determine significantly altered pathways informed by the translation changes at each time point for the two respective protein sources. mTOR pathway volcano plots used all of the molecules within the top 3 pathways ('EIF2 Signaling', 'Regulation of eIF4 and p70S6K Signaling', 'mTOR Signaling') in either cheese or milk for comparison, yielding presentation of the translation for 202 total transcripts, at 3 contrasts (60 vs. 0 min translation f.c.; 180 vs. 0 min translation f.c. and 180 vs. 60 min translation f.c.).

#### *2.7. Statistical Analyses*

Subject characteristics were compared between males and females using a t-test. Because there were no notable differences between males and females in major outcomes (i.e., blood amino acids), subjects were pooled and all comparisons (Plasma NEFA, Insulin, Amino Acids, Immunoblotting, and IHC colocalization) were analyzed using a 2-Way ANOVA with repeated measures for product and time. When appropriate after a significant interaction was detected, Sidak's multiple comparisons post-hoc test was used to identify differences from baseline or between protein products at a given time point. For all analyses, differences were considered statistically significant at *p* < 0.05. All statistical calculations and graphs were completed using GraphPad Prism (v8).

#### **3. Results**

#### *3.1. Subject Characteristics*

A total of 24 young adult participants completed both trials of this study. This was made up of 12 males and 12 females (Table 1). As expected, men had greater height, body weight, and had more lean mass than females (*p* < 0.05). The men also had less daily step activity than the females (*p* < 0.05). There were no differences between the sexes in age, BMI, fat mass, body fat % or daily protein intake.

#### *3.2. Blood Insulin, Glucose and Non-Esterified Fatty Acids*

Milk induced a rapid spike in insulin 20 min after ingestion (2-Way ANOVA: Time\*Product Interaction, *p* < 0.0001; Sidak's multiple comparisons test, Milk different from baseline and from cheese at 20 and 40 min, *p* < 0.0001) while cheese consumption did not significantly change insulin at any time point (Figure 2A). Blood glucose decreased at 60 min following ingestion of either product, but this decrease occurred earlier for milk (40 min; Time\*Product Interaction, *p* < 0.0001) and was significantly lower than cheese (Figure 2B). Similarly, NEFA levels decreased after ingestion of either Cheddar cheese or

milk, (Time\*Product Interaction, *p* < 0.0001), but this response was further decreased for milk compared to Cheddar cheese (Sidak's multiple comparisons test, Cheese vs. milk 40 min post, *p* = 0.023). Additionally, NEFA levels were significantly elevated in response to both protein sources by 300 min, in comparison to baseline NEFA values (Sidak's multiple comparisons test, Cheese: *p* = 0.003; Milk: *p* = 0.011) (Figure 2C).

**Figure 2.** Plasma (**A**) insulin (pmol/L), (**B**) glucose (mg/dL), and (**C**) non-esterified fatty acids (mmol/L) in the fasted state (0 min) and over a 300 min time period following the ingestion of either cheese (solid line) or milk (dotted line) in men and women (*n* = 24). Different from baseline (0 min) for milk (\*) and cheese (\$), *p* < 0.05. #, Different between groups at the specific time point, *p* < 0.05.

#### *3.3. Plasma Branched-Chain, Essential and Non-Essential Amino Acids*

Total branched-chain amino acids (BCAAs) increased with different kinetics in response to ingestion of the respective products (Time\*Product Interaction, *p* < 0.0001) (Figure 3A). After milk, BCAAs returned to baseline by 240 min post, and cheese maintained higher BCAA levels out to 270 min. Milk induced significantly higher BCAA levels than cheese from 20 to 60 min post ingestion, and decreased gradually towards baseline as cheese induced significantly higher BCAA in plasma than milk between 120 and 210 min (Sidak's, *p* < 0.05). Plasma leucine exhibited a similar response as total BCAAs (Time\*Product Interaction, *p* < 0.0001) (Figure 3B), with both products increasing leucine levels out to 210 min and with cheese elevating leucine levels slightly longer to 240 min (Sidak's, *p* < 0.05). The leucine response occurred to a greater magnitude for milk from 20 to 60 min while cheese induced higher leucine levels (vs. milk) from 120 to 180 min. Plasma isoleucine (Time\*Product Interaction, *p* < 0.0001) (Figure 3C) increased out to 160 min for milk while cheese increased isoleucine levels out to 240 min. Milk had a greater isoleucine response compared to cheese from 20 to 60 min while cheese had a greater plasma isoleucine response than milk from 120 to 210 min (Sidak's, *p* < 0.05). Plasma

valine (Time\*Product Interaction, *p* < 0.0001) (Figure 3D) increased over the 300 min time course for cheese and out to 270 min for milk. This response was greater for milk at 20–60 min while the cheese induced a greater valine level than milk from 120 to 210 min (Sidak's, *p* < 0.05). Total essential amino acids (EAA) increased above baseline for milk out to 180 min while cheese increased total EAA out to 300 min (Time\*Product Interaction, *p* < 0.0001) (Figure 3E). Plasma EAA were higher for milk from 20 to 60 min (compared to cheese) while EAA were higher for cheese from 120 to 210 min (vs. milk). Non-Essential amino acids (NEAA) (Figure 3F) increased above baseline for milk from 20 to 100 min while NEAA were elevated above baseline from 40 to 180 min for cheese (Time\*Product Interaction, *p* < 0.0001). Milk induced a greater NEAA response at 20–60 min while cheese induced a greater response than milk from 120 to 300 min (except at 270 min). Despite differences in amino acid kinetics between the products, the area under the curve over 5 h for total BCAA, leucine, isoleucine, valine, total EAA, and total NEAA were not different between cheese and milk products (Figure 3A–F).

**Figure 3.** Plasma levels (μmol/L) of (**A**) branched-chain amino acids (Total BCAA), (**B**) leucine, (**C**) isoleucine, (**D**) valine, (**E**) essential amino acids (Total EAA), and (**F**) non-essential amino acids (Total NEAA) in the fasted state (0 min) and over a 300 min time period following the ingestion of either cheese (solid line) or milk (dotted line) in men and women (*n* = 24). Different from baseline (0 min) for milk (\*) and cheese (\$), *p* < 0.05. #, Different between groups at the specific time point, mboxemphp < 0.05. Units are in micromolar (μM). Note: Total EAA (**E**) does not include the BCAAs.

#### *3.4. Muscle mTORC1 Signaling and Localization*

Phosphorylated p70S6K(Thr389) (Time\*Product Interaction, *p* = 0.0005) and phosphorylated rpS6(Ser240/244) (Time\*Product Interaction, *p* < 0.0001) increased above baseline and were increased to a greater extent for milk at 60 min post ingestion compared to

cheese (Sidak's multiple comparisons test, *p* < 0.0001 for p70S6K and rpS6K) (Figure 4A,B). Phosphorylated Akt(Ser473) was significantly elevated 60 min post ingestion as a result of cheese or milk with no difference between cheese and milk (2-Way ANOVA: Main Effect of Time, *p* = 0.0097) (Figure 4C). There were no significant differences detected for phosphorylated AS160(Ser588) or phosphorylated GSK-3β(Ser9) (Figure 4D,E). Figure 4F is representative immunoblotting images for the phosphorylated proteins.

**Figure 4.** Skeletal muscle protein expression using immunoblotting for (**A**) p70S6K1(Thr389), (**B**) rpS6(Ser240/244), (**C**) Akt(Ser473), (**D**) AS160(Ser588), and (**E**) GSK-3β(Ser9) in the fasted state (0 min) and at 60 and 180 min following the ingestion of either cheese (solid line) or milk (dotted line) in men and women. Panels (**A**,**B**) are data for *n* = 24 while for (**C**–**E**) only *n* = 8 (4 M, 4 F) were analyzed. Panel (**F**) are representative images of immunoblotting. Phosphorylated protein levels were normalized to Ponceau-S. Different from baseline (0 min) for milk (\*), *p* < 0.05. #, Different between groups at the specific time point, *p* < 0.05.

Using immunohistochemistry to fluorescently label and measure the spatial distribution of mTOR, we did not detect changes to the colocalization of mTOR with the lysosomal protein, LAMP2 (Figure 5A). However, mTOR colocalization with the sarcolemma (WGA) was different between groups at 60 and 180 min and increased at 180 min only after cheese ingestion (Time\*Product Interaction, *p* = 0.042; Sidak's multiple comparisons test, *p* = 0.003) (Figure 5B). Representative images for DAPI, WGA, mTOR, LAMP2 and the overlay are found in Figure 5C.

**Figure 5.** Skeletal muscle mTOR colocalization using immunohistochemistry. Panel (**A**) represents Mander's k1 mTOR-LAMP2 colocalization and (**B**) Mander's k2 mTOR-WGA colocalization at baseline (0 min) and at 60 and 180 min following the ingestion of either cheese (solid line) or milk (dotted line) in men and women (*n* = 24). Panel (**C**) are representative images using immunohistochemistry. Different from baseline (0 min) for cheese (\$), *p* < 0.05. #, Different between groups at the specific time point, *p* < 0.05.

#### *3.5. Ribosomal Profiling*

A subset of subjects' muscle samples (*n* = 4) was used for ribosomal profiling. Ribosomal profiling captures ribosome protected mRNA fragments to measure active translation of specific transcripts using RNA sequencing libraries. Both cheese and milk altered the same top 3 Canonical Pathways related to mTORC1 signaling (IPA: EIF2 Signaling, Regulation of eIF4 and p70S6K Signaling, mTOR Signaling) (Figure 6A) at both 60 and 180 min, while only milk activated glucose metabolism-related pathways (Glycolysis I, Gluconeogenesis I) 60 min post ingestion. Next, we created a volcano plot for the significantly altered transcripts from within the top 3 Canonical Pathways for cheese and milk respectively, representative of all translation changes under control of mTORC1 signaling. As a result, we demonstrated a significant and dramatic milk-induced (in comparison to cheese) translational response from 0 to 60 min for these mTORC1 mediated molecules (Figure 6B). This response for milk was reduced at 0–180 min after ingestion while cheese-induced translation of mTORC1 molecules was maintained at similar levels as was observed at 60 min (Figure 6C). Moreover, translation changes across the 60–180 min time period (Figure 6D) highlight the observation that stimulation of mTORC1 pathway is reduced at 180 min after milk ingestion but persists with cheese.

**Figure 6.** Skeletal muscle analysis of translated mRNAs assessed using ribosomal sequencing before and after cheese or milk ingestion in men and women (*n* = 4; 2 M, 2 F). (**A**) Canonical pathways identified by ingenuity pathway analysis for cheese and milk at 60 and 180 min and (**B**) Volcano plot of the change in translated mRNAs for cheese (orange squares) or milk (gray squares) from 0 to 60 min, (**C**) 0 to 180 min, and (**D**) 60 to 180 min post ingestion.

#### **4. Discussion**

Our current understanding of amino acid kinetics and subsequent skeletal muscle anabolism following protein intake has been informed by isolated protein sources (and often in liquid form) such as whey [33], casein [34], soy [5], and leucine-enriched EAAs [35]. Recently, the study of solid protein-enriched whole foods, of which the food matrix can greatly alter protein digestion and absorption kinetics and the subsequent muscle anabolic signature, is a valuable next step in studying the impact of dietary interventions on muscle health and disease [36]. The purpose of our study was to examine the response to 20 g of protein from Cheddar cheese on plasma amino acids, free fatty acids, insulin, and glucose and the subsequent skeletal muscle mTORC1 signaling and mRNA translational response. To better contextualize the results of Cheddar cheese ingestion with what is known in the field, we utilized a crossover design with comparison to milk, a highly studied protein source with a well-characterized absorption profile and muscle anabolic response [19,37]. The results from this study indicate that Cheddar cheese had a slow, yet persistent amino acid circulation appearance and subsequent skeletal muscle mTORC1 signaling and mRNA translation response when compared with the quick absorption and potent but short-lived mTORC1 stimulation induced by milk. At the studied dosage, Cheddar cheese did not induce a plasma insulinogenic or muscle glycemic response, a known effect of milk [19,38], suggesting Cheddar cheese may be an interesting food choice for dietary strategies geared to promote muscle protein anabolism yet requiring strict glycemic control.

The primary finding of this study was that consumption of Cheddar cheese (65 g) amounting to 20 g of protein promoted a delayed, yet sustained plasma amino acid concentration over 5 h, compared with the acute and potent appearance of circulating amino acids induced by milk proteins. Even though the amounts of protein and leucine were similar between the products, milk resulted in a more rapid and robust amino acid response likely driven by the whey protein component (20% whey, 80% casein in bovine milk vs. 100% casein in Cheddar cheese). Nonetheless, we found it interesting that there was no difference in the total circulating amino acids across the entire 5 h time course between cheese and milk. Thus, although the solid food matrix of cheese and protein composition may slow the digestion and absorption of protein and, subsequently, amino acid release into the circulation, when matched for protein, cheese and milk have similar total plasma amino acid availability. Casein hydrolysate, the form present in Cheddar cheese, has shown to result in a greater appearance in circulating levels of leucine compared with intact casein [4,17,39]. Though difficult to compare to a liquid casein beverage, the plasma leucine appearance data following 20 g Cheddar cheese protein ingestion demonstrated a slower plasma leucine appearance rate and magnitude compared to a similar amount of isolated casein hydrolysate [4] suggesting that the complex matrix of cheese may delay the release of amino acids into circulation [40]. It is currently unknown if a longer aged Cheddar cheese may speed the circulating appearance of amino acids. However, when compared with other solid, protein-dense foods, such as pork [6], cooked egg [14], and steak [13], Cheddar cheese aged to one month produced a similar plasma appearance, magnitude, and sustained amino acid availability response.

We also measured muscle mTORC1 activation using three different approaches with the cumulative result of these assessments demonstrating that anabolic signaling tracked closely with circulating amino acids for each product and with milk demonstrating a more robust mTORC1 signaling response early after intake (1 h). This is logical since essential amino acids, especially leucine, along with insulin, which also peaked prior to 1 h, are stimulators of mTORC1-mediated protein synthesis [41,42]. It is likely that the insulin response from milk, combined with the quickly absorbed leucine, synergized to enhance mTORC1 signaling as noted by the magnitude of p70S6K and rpS6 activation [42]. It is well known that anabolic cues such as insulin, mechanical stimulation, and amino acid ingestion stimulate mTORC1 and its downstream effectors (e.g., S6K1) to enhance translation initiation [43]. While our measurement of mTORC1 signaling was limited to a 3 h time course (based on other protein-dense whole food studies [6,13,14]), we found it noteworthy that mTORC1 activation following Cheddar cheese ingestion persisted at 3 h (and possibly beyond) in accordance with the plasma amino acid appearance and as supported by the mTORC1 localization data and the ribosomal profiling of translated mRNAs under control of mTORC1 signaling. It is unclear how a sustained circulation of amino acids following Cheddar cheese intake may impact muscle protein accretion. Whole foods that are slow digesting (in comparison to commonly studied dairy protein drinks) may have utility in sustaining the free amino acid pool so that they have a longer window to synergize with other anabolic cues such as exercise, or by offsetting protein breakdown to enhance net protein balance when combined with the acute stimulus of a faster digesting protein source [3,44]. For example, drinking a small glass of milk with cheese, may result in a greater net protein balance over several hours in comparison to a bolus of milk alone, because of the ceiling for acute anabolic activation and subsequent oxidation of excess amino acids (coined the 'muscle-full effect') thereby limiting the anabolic benefit of the beverage [38,44,45]. There is a similar underlying premise behind the recommendation of ingesting casein (a major component of cheese) prior to bedtime to enhance exercise adaptations [46–48].

Another interesting observation about the acute response to Cheddar cheese intake (in contrast to milk) in this study, was that cheese did not observably increase circulating insulin or translation of muscle mRNAs related to glycolytic pathways at any time point we measured after ingestion. Therefore, if a dietary intervention requires strict glycemic control, such as for individuals with diabetes [49] or requires adhering to a ketogenic diet [50], cheese may be a valuable protein food source to keep on the menu. The two most likely reasons for why cheese and milk ingestion had different circulating insulin responses, are that (a) milk contains higher levels of carbohydrates and (b) milk induced an early spike in circulating serum leucine (compared with cheese), which stimulates endogenous insulin release [51]. In addition to being less glycemic, the general public's health opinion of cheese should be re-examined since regular consumption of cheese does not appear to influence LDL or HDL levels despite the characteristically high fat content [52]. Though Cheddar cheese does incorporate a significant portion of its calories from fat, fat does not appear to influence muscle protein anabolism [53], and may even synergize with protein to promote a greater anabolic response [14,36]. However, because of the extra calories associated with fat as compared to other macronutrients, individuals who must restrict their calories may benefit from reduced fat cheese.

#### **5. Conclusions**

In summary, Cheddar cheese provided a slow and sustained appearance of circulating amino acids and subsequent activation of mTORC1 signaling when compared to milk matched for protein (and leucine) content. Also, Cheddar cheese at the amount consumed in this study did not noticeably increase circulating insulin or induce a muscle glycemic response in contrast with milk. Overall, low fat Cheddar cheese should be considered as a protein-dense food choice given its high leucine content, ability to sustain amino acid levels and promote protein anabolism and, especially, considering its low glycemic properties. Future studies are needed to examine muscle protein accretion in response to daily Cheddar cheese ingestion when combined with habitual exercise.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-664 3/13/2/614/s1, Table S1: Amino acid composition of Cheddar cheese and milk.

**Author Contributions:** Conceptualization: E.D.B., L.S.W., M.J.D.; Methodology: M.J.D., P.T.R., M.J.B., E.D.B., L.S.W., M.T.H.; Clinical Execution and Data Collection: N.M.M.P.d.H., Z.S.M., P.T.R., J.J.K., A.I.M., J.J.P., L.M.B.; Writing: N.M.M.P.d.H., Z.S.M., M.J.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** Funding for this study was provided by BUILD Dairy and Glanbia Nutritionals, partial clinical support by the National Center for Advancing Translational Sciences of the National Institutes of Health (UL1TR002538), and postdoctoral support from the Ruth L. Kirschstein National Research Service Award NIH 1T32HL139451.

**Institutional Review Board Statement:** Enrolled participants read and signed the informed consent document, which was approved by the University of Utah Institutional Review Board (IRB #110963) and in agreement with the Declaration of Helsinki.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Raw sequence data can be obtained from the National Center for Biotechnology Information Gene Expression Omnibus repository entry GSE163279.

**Acknowledgments:** We would like to thank J. Alan Maschek, University of Utah Metabolomics, Proteomics and Mass Spectrometry Cores, 15 N Medical Drive East, UT 84112, USA, for his contributions. We would also like to thank the clinical services core for their assistance with the blood sampling and muscle biopsies.

**Conflicts of Interest:** E.D.B. and L.S.W. are employed by DairyWest and Glanbia Nutritionals, respectively.

#### **References**

1. Atherton, P.J.; Etheridge, T.; Watt, P.W.; Wilkinson, D.; Selby, A.; Rankin, D.; Smith, K.; Rennie, M.J. Muscle full effect after oral protein: Time-dependent concordance and discordance between human muscle protein synthesis and mTORC1 signaling. *Am. J. Clin. Nutr.* **2010**, *92*, 1080–1088. [CrossRef]


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Nutrients* Editorial Office E-mail: nutrients@mdpi.com www.mdpi.com/journal/nutrients

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-4940-8