Effects of Whole Milk Supplementation on Gut Microbiota and Cardiometabolic Biomarkers in Subjects with and without Lactose Malabsorption

The aim of this study was to compare the impact of whole milk supplementation on gut microbiota and cardiometabolic biomarkers between lactose malabsorbers (LM) and absorbers (LA). We performed a pair-wise intervention study of 31 LM and 31 LA, 1:1 matched by age, sex, body mass index, and daily dairy intake. Subjects were required to add 250 mL/day whole milk for four weeks in their routine diet. At the beginning and the end of the intervention period, we collected data on gut microbiota and cardiometabolic biomarkers. Whole milk supplementation significantly increased Actinobacteria (P < 0.01), Bifidobacterium (P < 0.01), Anaerostipe (P < 0.01), and Blautia (P = 0.04), and decreased Megamonas (P = 0.04) in LM, but not LA. Microbial richness and diversity were not affected. The fecal levels of short-chain fatty acids (SCFAs) remained stable throughout the study. Body fat mass (P < 0.01) and body fat percentage (P < 0.01) reduced in both groups, but the changes did not differ between groups. No significant differences in other cardiometabolic markers were found between LM and LA. When compared with LA, whole milk supplementation could alter the intestinal microbiota composition in LM, without significant changes in fecal SCFAs and cardiometabolic biomarkers.


Introduction
Milk has been commonly regarded as an essential component of a balanced diet, since it is rich in high-quality protein, calcium, as well as plenty of health-promoting bioactive components, such as whey peptides, conjugated linoleum acid, oligosaccharides, and immunoglobulin [1]. Previous prospective studies suggested that milk consumption was significantly associated with a lower risk of obesity [2], hypertension [3], type 2 diabetes [4], and cardiovascular disease [5]. In humans, the ability to digest lactose, which is the main carbohydrate present in milk, depends on

Materials and Methods
Prior to initiating the study, the study design and consent form were reviewed and approved by the Ethics Committee of the School of Public Health, Tongji Medical College, Huazhong University of Science and Technology (approval No. 12012015). The study was registered at clinicaltrials.gov (NCT02798718).

Study Participants
This study was a 1:1 pair-wise matching intervention study in lactose absorbers and malabsorbers. At the beginning, 233 participants were voluntarily recruited from Tongji Medical College of Huazhong University of Science and Technology. The inclusion criteria included healthy volunteers aged ≥18 years, Han population, and average dairy intake less than one serving during the past year. We then excluded the participants with a history of acute or chronic gastrointestinal disorders, any known metabolic disease (diagnosed diabetes, hypertension, or cardiovascular disease), consumption of antibiotics or probiotics within the preceding month, use of medications or dietary supplements that could influence glucose or lipids metabolism, pregnancy or lactation, excessive alcohol consumption, significant body weight variation in the past three months, or a known allergy to milk.
Then, 227 eligible participants underwent a hydrogen breath test after lactose administration to examine the presence of lactose malabsorption. Written informed consent was obtained from all the participants before participating. This test was performed with a hydrogen breath test analyzer (Shenzhen Zhonghe Headway Bio-Sci & Tech Co., Ltd., Shenzhen, China). Briefly, after a standardized low-carbohydrate dinner and a 12-h fast, the end-expiratory breath H 2 was determined before and at 30-min intervals after a 25-g lactose load for the ensuing 3 h. Eating, smoking, and exercising were Nutrients 2018, 10, 1403 3 of 14 not allowed throughout the test. A rise of 20 ppm in breath H 2 over basal values was considered as an indication of lactose malabsorption [17]. Finally, 31 LA and 31 LM were recruited by 1:1 pair-wise matching with age (±3 years), sex, body mass index (BMI) (±10%), and the average dairy intake (<125 g/day, 125-250 g/day) for the intervention study. The details of participant recruitment are shown in Supplementary Figure S1.

Milk Intervention and Dietary Intake Assessment
In one-week run-in phase, all of the participants were required to maintain their initial diet. In the following four-week intervention, they were further instructed to add extra one box of 250 mL whole milk (Mengniu Dairy (Group) Co., Ltd., Hohhot, China) in their routine diet per day. All of the subjects were instructed to maintain their usual physical activity throughout the entire study. The participants were asked to return the empty milk package to assure compliance.
To quantify macronutrient composition, individuals were instructed to complete a three-day food record (including two weekdays and one weekend day) at the end of run-in and intervention periods. Before start of the study, the participants were instructed by investigators how to weigh and record their food intake. The dietary records were checked, in the case of missing data, and analyzed by the investigators. Energy and nutrient intake were calculated according to the "Chinese Food Composition Tables" [18].

Cardiometabolic Biomarkers Measurement
Anthropometric and biochemical measurements were conducted at the beginning and end of the intervention period. Body composition was measured by bioelectric impedance analysis using Inbody 720 (BIOSPACE CHINA Inc., Shanghai, China). Body weight (kg) and body height (cm), were measured with standardized techniques. BMI (kg/m 2 ) was calculated as body weight divided by the square of height. Waist circumference, hip circumference, and blood pressure were measured with standardized techniques.
Fasting blood samples were drawn from an antecubital vein into heparinized tubes in the morning after a 12-h fast. Plasma was separated in a 4 • C centrifuge and stored at −80 • C until analysis. Fasting plasma glucose (FPG), total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were analyzed by colorimetric enzymatic methods using commercial kits (BioSino Bio-Technology & Science Inc., Beijing, China). Fasting plasma insulin (FPI) and C-peptide were analyzed with ELISA kits (Mercodia AB, Uppsala, Sweden). C-reactive protein (CRP) was determined with ELISA kits (R&D Systems, Inc., Minneapolis, MN, USA). Malondialdehyde (MDA) was measured according to the thiobarbituric acid method with commercial kits (Jiancheng Bioengineering Institute, Nanjing, China). Intraassay variation for FPG, TC, TG, LDL-C, HDL-C, FPI, C-peptide, CRP, and MDA ranged from 0.9% to 5.5% and interassay variation ranged from 1.2% to 6.5%. Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated from fasting glucose and insulin while using the following formula [19]:

Fecal Samples Collection and DNA Extraction
Participants were instructed to collect their stool using aseptic swabs and were sealed dung cup before and after the intervention study. Uncontaminated samples were collected and immediately stored in a provided cooler, then frozen at −80 • C within 30 min. DNA extraction was performed using a QIAamp Fast DNA Stool Mini Kit (Qiagen, Valencia, CA, USA). The concentration of bacterial DNA was measured using Nanodrop 2000 (Thermo Scientific, Wilmington, NC, USA).

Fecal Microbiota Analysis
The 16S rRNA sequencing data were processed using the Quantitative Insights Into Microbial Ecology platform (QIIME, Boulder, CO, USA, V.1.7.1) [20]. Raw sequencing reads were demultiplexed and filtered. Chimeras were removed and the operational taxonomic units (OTUs) were with 97% homology were generated using USEARCH [21]. The identified taxonomy was aligned with Ribosomal Database Project Classifier [22] and Greengenes reference database (V.13.8). To ensure an even sampling depth, all of the samples were rarefied to the lowest read number. Alpha and beta diversity metrics were calculated in QIIME using the rarefied OTU table. Beta diversity was estimated using weighted and unweighted UniFrac distances between samples and was visualized with principal coordinate analysis (PCoA).

Fecal SCFA Analysis
Fecal samples (0.5 g) were homogenized after addition of 5 mL of ultrapure water and centrifuged at 4000 rpm for 10 min at 4 • C. The supernatant fluid (600 µL) was mixed with 120 µL 25% phosphoric acid, homogenized and centrifuged at 12,000 rpm for 10 min at 4 • C. Then, the supernatant was filtered through a Millex-GS 0.22-mm syringe filter unit (Millipore, Burlington, MA, USA). The filtrate was mixed with 2-ethylbutyrate (DRE, Altshausen, Germany) as the internal standard with 1 mM. A mixed-SCFA standard solution was prepared by using analytical quality (>99% purity) reagents (Sigma-Aldrich, Bornem, Belgium). SCFAs were quantified with an Agilent 6890N gas chromatograph coupled with an Agilent 5975B mass spectrometer (Agilent Technologies Santa Clara, CA, USA). The capillary GC column was an Agilent HP-INNOWAX 30 m × 0.25 mm, 0.25 µm film thickness, with helium as the carrier gas at a constant flow rate of 2.8 mL/min. The GC oven temperature program was as follows: initial temperature 72 • C, then to 170 • C at 13 • C/min, to a final temperature 230 • C, where it was held for 5 min, at 30 • C/min. The temperatures of the injector, transfer line, and detector were set at 200 • C, 280 • C, and 220 • C, respectively. The total run time was 14.54 min. Acetate, propionate, and butyrate were quantified with appropriate calibration curves obtained using internal standard quantitation. The Varian MS workstation software (version 6.6) was used for data acquisition and processing. The interday and intraday coefficient of variances ranged from 4.6% to 8.7%.

Quantitative Polymerase Chain Reaction (qPCR)
PCR amplification and detection were conducted in 96-well plates with PrimeScript TM RT reagent kit (TAKARA BIO INC., Dalian, China) on a DNA Engine Opticon 2 fluorescence detection system (BIO-RID, Hercules, CA, USA). Samples were run with a final volume of 25 µL, containing 0.4 µM of each primer and 2 µL of the respective template DNA. The specific primers for Bifidobacterium forward CTCCTGGAAACGGGTGG and reverse GGTGTTCTTCCCGATATCTACA [23]. Amplifications were done with the following temperature profiles: one cycle at 95 • C (30 s), 40 cycles of denaturation at 95 • C (5 s), 61 • C (30 s), and 72 • C (45 s), and a final one cycle of 94 • C (5 min). Quantification was done by using standard curves made from know concentrations of plasmid DNA, which was prepared according to Sabine et al. [24].

Statistical Analysis
All the data were expressed as the mean ± standard error of the mean (SEM). Dietary energy and nutrients intake were analyzed using Student's paired t-test. For cardiometabolic biomarkers, baseline differences were evaluated using a Student's unpaired t-test. Between-group differences were analyzed using a two-factor repeated measures analysis of variance (ANOVA) with time (pre, post) and group (LM, LA) as the factors. In the case of a significant time effect being shown, post-hoc analyses with Bonferroni correction were applied to identify significant within-group differences. The Kolomogorov-Smirnov test was performed to test the normality of distribution. If skewed distribution, data was logarithmically transformed. Statistics were performed with SPSS 24.0 (IBM, Brussels, Belgium).
For phylum-or genus-level microbiota groups, comparisons between two groups were evaluated with Mann-Whitney tests, within-group differences were analyzed using Wilcoxon signed-rank tests, between-group differences were evaluated using a linear mixed model with package "lme4" taking into account of repeated measurements, group, gender, and BMI. Data of qPCR were expressed as logarithmically transformed. ANOVA with Turkey post-hoc analysis was applied to compare the alpha diversity between and within groups. Beta diversity was assessed based on unweighted and weighted UniFrac distances and its significance was determined using permutation-based ANOVA with the use of the Adonis function in the "vegan" package. P values were corrected for multiple comparisons using the Benjamini-Hochberg procedure. Enterotyping on the basis of Partitioning Around Medoids with Jensen-Shannon divergence was conducted on a combined genus-abundance matrix of all samples [25]. Besides, the optimal number of cluster was determined by the Calinski-Harabasz Index. Statistics were performed using R version 3.3.3. A two-tailed P < 0.05 was considered to be statistically significant. A corrected false discovery rate (FDR) < 0.2 was considered as significant [26].

Results
A total of 62 subjects, 31 LA and 31 LM, participated in the intervention study. There were 44 men and 18 women with an average age of 24.7 ± 0.3 years and an average body mass index of 22.0 ± 0.4 kg/m 2 . No subjects dropped out of the study during the intervention period (See Supplementary Figure S1). No significant differences were found between LA and LM at the baseline (Table 1). No adverse events and gastrointestinal discomforts were reported during the entire study. Energy and macronutrient intake, as assessed by self-reported three-day food records, remained stable throughout the study, but calcium intake significantly increased in all of the participants following the whole milk intervention (See Supplementary Table S1).

Fecal Short-Chain Fatty Acids Concentrations
Fecal acetate, propionate, and butyrate concentration did not significantly differ within and between the groups (Figure 4).  Box represents the interquartile range, the line inside represents the median, whiskers represent 10-90 percentiles, "+" represents outliers that are past the ends of the whiskers. (c) Sample distribution over enterotypes before and after whole milk supplementation in lactose malabsorbers (LM) and absorbers (LA). (d) Comparisons of changes in the microbiota abundance in LM and LA with different enterotypes (LM with Bacteroides enterotype, n = 24; LA with Bacteroides enterotype, n = 18; LM with Prevotella enterotype, n = 7; LA with Prevotella enterotype, n = 11). Data are presented as Mean ± SEM (*, P < 0.05, and **, P < 0.01).

Fecal Short-Chain Fatty Acids Concentrations
Fecal acetate, propionate, and butyrate concentration did not significantly differ within and between the groups (Figure 4).

Body Composition and Cardiometabolic Markers
Although body weight and BMI remained unchanged during the intervention study, body fat mass (P = 0.03 for LM; P < 0.01 for LA) and the proportion of body fat (P = 0.02 for LM; P < 0.01 for LA) significantly decreased in both groups ( Table 2). Changes in body composition did not differ between LM and LA. No significant within-or between-group differences were observed in blood pressure, FPG, FPI, C-peptide, TG, TC, LDL-C, HDL-C, CRP, and MDA. Change in LDL-C was inversely associated with a change in Blautia abundances among LM (P = 0.01), but no shifts in the cardiometabolic biomarkers were associated with alternations in the relative abundance of Bifidobacterium, Anaerostipes, and Megamonas (See Supplementary Tables S3-S6). Table 2. Body composition and cardiometabolic biomarkers before and after whole milk supplementation in subjects with or without lactose malabsorption 1 .

Discussion
In the present study, as compared with LA, four-week 250 mL/day whole milk supplementation selectively altered gut microbiota composition in LM, especially in those with Bacteroides enterotype, without significant effects on the overall microbiota richness and diversity. Additionally, no significant differences were found in alternations in fecal SCFAs or cardiometabolic biomarkers between LM and LA.
We observed that four-week whole milk supplementation significantly increased the relative abundances of Bifidobacterium in LM, which was mainly consistent with a previous study, showing the bifidogenic effects of lactose in healthy LM in western population [13]. Bifidobacterium could utilize lactose as a preferable energy source rather than glucose [27], and thus the abundance of Bifidobacterium could increase with higher lactose intake derived from milk supplementation [28]. Bifidobacteria is believed to play a pivotal role in gut microbiota homeostasis and human health [29]. Several strains in bifidobacteria have been widely used as probiotics because of both direct and indirect influences on enhancing immunity, inhibiting the growth of pathogenic bacteria, and treating inflammatory disease [30]. In our study, the abundance of Anaerostipes and Blautia also increased with whole milk supplementation in LM. Anaerostipes was a butyrate-production bacteria, and the cross-feeding relationship between Anaerostipes and Bifidobacterium might account for its alteration [31]. Blautia was acetogenic and might benefit from the production of hydrogen, a production from lactose fermentation by Bifidobacterium [32]. Additionally, our results indicated that whole milk supplementation induced different alternations in microbiota composition across different enterotypes, which was in line with previous findings that individuals with distinct enterotypes differed in responses to the same diet intervention [33,34].
There were no significant changes in the fecal levels of SCFAs following four-week supplementation of whole milk, despite the significant changes in several SCFA-producing microbiota taxa. One potential explanation for this might be conversion among SCFAs through microbial cross-feeding, in particular, from acetate to butyrate [35], which was mainly metabolized in the colon to supply energy for colonocytes [36]. But, some studies suggested that fecal butyrate significantly increased following 8-week intervention of undigested carbohydrate intervention, but did not following four-week intervention [15]. Therefore, relative short intervention duration might be another reason accounting for the unchanged fecal concentrations of SCFAs in this study, and long-term studies are needed to further verify this finding in the future.
In this study, whole milk supplementation did not significantly affect the cardiometabolic biomarkers, except slight decreases of body fat mass and body fat percentage in LM and LA. Several intervention studies suggested that consumption of dairy products might have a beneficial effect on cardiometabolic factors [37][38][39][40]. A meta-analysis of randomized controlled trial elucidated that dairy products reduced body fat mass in short-term intervention studies, but not in long-term [37], which well supported the results of our study. A possible explanation might be the differential compliance with the intervention protocol between short and long trials [41].
In our study, the marked microbial changes in LM did not affect carbiometabolic biomarkers. Previous studies showed that Bifidobacteria improved diabetes induced by high-fat diet [29,42] but we did not observe any association between Bifidobacterium and carbiometabolic biomarkers. When compared the Bifidobacterium of our study population with that of other population, we found the baseline abundance of Bifidobacterium was markedly lower in our study population, but it reached slightly higher abundance after whole milk population as compared to this healthy, untreated population [26,43]. Therefore, one reason for lack of effects on the cardiometabolic biomarkers in LM may be the significant but small increase as well as inconsistent pattern of increase. Our observations indicated that moderate whole milk supplementation (250 mL/day) had no adverse cardiometabolic effect in LM, compared with LA. Accordingly, it is possibly acceptable for LM to consume moderate whole milk per day for health promotion, which would draw more attention in the countries of high prevalence of lactose malabsorption.
There are several strengths of the current study. To our best knowledge, no previous study has comprehensively compared the effects of whole milk supplementation on the gut microbiota composition and its crucial metabolites (SCFAs) and cardiometabolic biomarkers in LM and LA. We also investigated the possible different responses to whole milk in subjects with different enterotypes. Our study also has certain limitations. Firstly, we did not include a group without whole milk intervention, so the potential effects of non-treatment factors could not be excluded. However, the 1:1 pair-wise design allowed us to evaluate the different effects of whole milk supplementation between LM and LA. Secondly, we did not control the diet of the participants, but added whole milk to the initial diet. However, total energy and macronutrients intake was not different between two groups, and the effects of changes in diet could be excluded. Thirdly, four weeks may be relative short to observe changes in cardiometabolic biomarkers of healthy adults aged from 20-30 in this study, whose cardiometabolic biomarkers maintain a relatively steady state. Long-term studies or studies in the participants of different metabolic statuses are needed to replicate our findings and precisely assess the effect of milk on cardiometabolic outcomes.

Conclusions
In conclusion, when compared with LA, 250 mL/day whole milk supplementation could alter the intestinal microbiota composition in LM, without alternations in fecal SCFAs and metabolic biomarkers. It may be acceptable to include moderate milk as a part of the routine diet of LM, as a suggestion to personalize milk consumption, and thereby enhance the levels of dietary calcium intake and improve health. Further studies are warranted to confirm our findings and clarify the underlying mechanisms.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/10/10/1403/ s1, Figure S1: Flow chart of the participant recruitment and withdraw, Table S1: Dietary energy and nutrients intake based on a three-day food records at baseline and following 4-week supplementation of whole milk, Table  S2: Comparison of gut microbiota composition at the phylum and genus level betweem lactose malabsorbers (LM) and absorbers (LA), Table S3: Comparison of changes in cardiometabolic biomarkers between groups based on changes in Bifidobacterium abundance among lactose malabsorbers, Table S4: Comparison of changes in cardiometabolic biomarkers between groups based on changes in Anaerostipes abundance among lactose malabsorbers, Table S5: Comparison of changes in cardiometabolic biomarkers between groups based on changes in Blautia abundance among lactose malabsorbers, Table S6: Comparison of changes in cardiometabolic biomarkers between groups based on changes in Megamonas abundance among lactose malaborbers.