Characteristic chromatographic fingerprint study of short-chain fatty acids in human milk, infant formula, pure milk and fermented milk by gas chromatography-mass spectrometry

Human milk, infant formula, pure milk and fermented milk as food products or dietary supplements provide a range of nutrients required to both infants and adults. Recently, a growing body of evidence has revealed the beneficial roles of short-chain fatty acids (SCFAs), a subset of fatty acids produced from the fermentation of dietary fibers by gut microbiota. The objective of this study was to establish a chromatographic fingerprint technique to investigate SCFAs in human milk and dairy products by gas chromatography coupled with mass spectrometry. The multivariate method for principal component analysis assessed differences between milk types. Human milk, infant formula, pure milk and fermented milk were grouped independently, mainly because of differences in formic acid, acetic acid, propionic acid and hexanoic acid levels. This method will be important for the assessment of SCFAs in human milk and various dairy products.


Introduction
Milk and its derivatives have been proposed as useful foods throughout all stages of life (Visioli & Strata 2014), and during infancy, childhood and adolescence in particular. As the gold standard of infant nutrition (da Costa Ribeiro et al. 2015;Wernimont et al. 2015), human milk contains the optimal composition of nutrients to meet the nutritional need of the newborn (Zhang et al. 2014;Ren et al. 2015). The World Health Organization (WHO), American Academy of Pediatrics (AAP) and numerous other health organizations recommend exclusive breastfeeding up to six months of life (Gartner et al. 2005;Basch et al. 2013). However; infants often cannot be breastfed for a range of reasons (Koletzko et al. 2005;Kent et al. 2015), and infant formula provides an alternative source of nutrition. Milk and milk products (including pure milk and fermented milk) are nutritious food items rich in numerous nutritional components (Haug et al. 2007;Dixit et al. 2015), including macronutrients and micronutrients (Lipkie et al. 2015). The consumption of milk daily could supply a significant amount of the recommended nutrients that are required daily (Haug et al. 2007). For centuries, milk and dairy products have been widely consumed in the human diet (Suturović et al. 2014), especially in childhood and adolescence. In conclusion, milk and dairy products are one of the most important foods sources (Suturović et al. 2014).
In addition to providing nutrition, human milk and dairy products also provide probiotics and/or prebiotics that assist the development, maturation, modulation and homeostasis of gut microbiota (Ward et al. 2006;Gueimonde et al. 2007;Ceapa et al. 2013;Fan et al. 2014;da Costa Ribeiro et al. 2015;Gagnon et al. 2015). The human intestinal tract harbors 10 13 to 10 14 symbiotic microorganisms (Gill et al. 2006), which are collectively called the gut microbiota. Although previously described as a ''forgotten organ'' (O'Hara & Shanahan 2006), gut microbiota play an essential role in human health, including systemic homeostasis, immune regulation, nutrient metabolism and protection against pathogens (Hooper & Gordon 2001;Sekirov et al. 2010;Tlaskalov a-Hogenov a et al. 2011;Young 2012). The end products of gut microbial fermentation (den Besten et al. 2013;Kim et al. 2013), short-chain fatty acids (SCFAs), serve as messengers that reveal the mutually beneficial relationship between gut microbiota and the host (Vinolo et al. 2011;Ganapathy et al. 2013). SCFAs are volatile fatty acids with between one and six carbon atoms, which commonly include formic, acetic, propionic, butyric, isobutyric, valeric, isovaleric and hexanoic acid (Bergman 1990). In the colon, SCFAs have a positive effect on intestine and gut microbiota, specifically, SCFAs are used as the major energy source for colonic epithelial cells (Topping & Clifton 2001;Hamer et al. 2012) and help to regulate gene expression, cell proliferation, differentiation and apoptosis (Topping & Clifton 2001;Tuohy et al. 2012). SCFAs not only serve as a carbon source for gut microbiota (Fischbach & Sonnenburg 2011), but also prevent the overgrowth of pathogenic bacteria (Proh aszka et al. 1990). In order to travel to tissues and organs beyond the colon, SCFAs are taken up from the intestinal lumen, mainly by passive diffusion (non-ionic) and active transport (SCFA --HCO 3 exchange, MCT and SMCT1 transporters) processes (den Besten et al. 2013). The beneficial effects of SCFAs on human health have been well documented (Licciardi et al. 2011;Vinolo et al. 2011;den Besten et al. 2013;Kim et al. 2014;Kasubuchi et al. 2015), and demonstrated as acting via regulating at least four signaling mechanisms (HDACs, FFAR2/FFAR3, AMPK and K-cAMP) (Li et al. 2014).
In this work, we aimed to investigate SCFA levels in human milk and dairy products. Accordingly, a rapid and reliable gas chromatography-mass spectrometry (GC-MS) technique along with a characteristic chromatographic fingerprint method were combined to semi-quantify SCFAs in 45 human milk samples from six healthy lactating mothers, five mainstream-brand infant formula samples, 12 pure cow milk samples and eight fermented milk samples. Subsequently, semiquantitative results were processed by unsupervised pattern recognition using PCA, which indicated that the content of SCFAs in human milk, infant formula and dairy products are significantly different. This study provides a valuable reference for the further development of protocols that assess the SCFAs of human milk and various dairy products.

Sample collection
In this work, 25 batches of dairy products including five samples of infant formula (IF 1-5), 12 samples of pure milk (PM 1-12; commercial name, ultra-high temperature sterilized milk) and eight samples of fermented milk (FM 1-8) were purchased from different brands and manufacturers in local supermarkets. All the samples were stored at 4 C until analysis. The details of all samples are recorded in Table S1.
Human milk samples were provided by six healthy lactating mothers aged between 27 and 34 years, and written informed consent was obtained from each of these participants. Forty-five samples (HM-1-1-HM-1-8 to HM-6-1-HM-6-8, Table S2) were expressed during the first eight weeks post-partum and transferred directly into 5-mL Eppendorf tubes by the mothers. The collected milk was immediately stored at À20 C and then transferred to the À80 C freezer until analysis. The sampling procedures applied ensured that breastfeeding had been well established and that the baby was thriving.

Sample preparation
For milk samples (human milk, pure milk and fermented milk), a 100 lL sample was aliquotted and mixed with 300 lL 0.5% (v/v) hydrochloric acid ethanol solution (containing 2.64 lg/mL IS) to precipitate the protein. All samples were vortexed for 1 min and centrifuged at 14,000 rpm for 10 min at 4 C.
For the infant formula samples, accurately weighed powder (0.5 g) was immersed in 5 mL 0.5% (v/v) hydrochloric acid ethanol solution (containing 2.64 lg/ mL IS), ultrasonically extracted for 5 min. The extracted solution was centrifuged at 14,000 rpm for 10 min at 4 C. The supernatants were transferred to the sample vial and stored on ice until GC-MS analysis.

GC-MS condition
Chromatographic analysis was performed on a GC-2010 Plus (Shimadzu, Japan) equipped with an automatic injection system (AOC-20i, Shimadzu, Kyoto, Japan) and a split/splitless injection port with a glass liner packed with glass wool (Restek). Chromatographic separation was achieved on a DB-FFAP capillary column (30 m Â 0.25 mm Â 0.25 lm, Agilent). The temperature of the inlet was 250 C and the volume of injection was 5 lL. The temperature program was as follows: the initial oven temperature at 50 C held for 1 min, ramped to 180 C at 10 C/min. The carrier gas was helium with a constant linear velocity at 1 mL/min. The import of the sample into the column was accomplished in the split mode at 10:1.
MS analysis was performed using a GCMS-QP2010 Ultra (Shimadzu, Japan) equipped with an electronic impact (EI) source. MS conditions were optimized as follows: electron energy, À70 eV; ion source temperature, 250 C; interface temperature, 250 C; solvent delay time, 4 min; and detection voltage, 1.25 kV. The scan mode (mass range, 33-300 amu) was implemented to identify SCFAs and develop a selected ion monitoring (SIM) method. The SIM mode was used to generate a characteristic chromatographic fingerprint of SCFAs in human milk and dairy products. The optimal SIM parameters of SCFAs were shown in Table 1.

Data processing
Principal component analysis (PCA) was performed using EZinfo software (Waters, Milford). The Pareto distribution as a scale for variables was used for the multivariate statistical analysis. Boxplots were charted by Origin 8 software (OriginLab Co., Northampton, MA).

Identification of the six target SCFAs in milk samples
Six SCFAs were unambiguously identified by comparing the retention time and diagnostic fragment ions with those of true standards. The representative TIC profile of human milk is shown in Figure S1. The fragmentation pathway of hexanoic acid is shown in Figure S2, as a model to illustrate the typical fragmentation pattern of SCFAs.
The EI source, a type of hard ionization source, generally yields a low abundance of molecular ions that are not suitable for the accurate determination of SCFAs. Thus, the three highest intensity ions were selected to develop the SIM method. Specifically, the ion with the highest abundance was set as channel 1 for the semi-quantitation step of SCFAs, and other ions were set as channels 2 and 3 for the qualitative diagnosis of compounds.

Optimization of sample preparation and chromatographic conditions
SCFAs, particularly formic acid, acetic acid and propionic acid, can ionize in polar solvent, which reduces the amount of SCFAs in the vaporization chamber. In order to inhibit the ionization of SCFAs during the process of sample preparation, ethanol was used instead of methanol and hydrochloric acid was added into the ethanol. By comparing ethanol containing 0.01%, 0.5% or 1% (v/v) hydrochloric acid, 0.5% (v/v) hydrochloric acid ethanol solution could significantly suppress the ionization of SCFAs with low levels of corrosiveness to the instrument. Furthermore, in order to improve the sensitivity and peak shape of SCFAs in the characteristic chromatographic fingerprint, the injection volume was set at 5 lL and the split ratio was 10:1.

Method validation
Precision, repeatability and stability were assessed to validate the characteristic chromatographic fingerprint method, and the results are summarized in Table 2. Intra-and inter-day variations were used to investigate the precision of the characteristic chromatographic fingerprint method by analyzing a single sample six times on 1 day and 3 consecutive days, respectively. All relative standard deviations (RSDs) of the precision values did not exceed 5.04%. The repeatability assay was evaluated by analyzing six parallel samples twice, with RSDs <4.42%. These results demonstrated that the established characteristic chromatographic fingerprint method was reliable and reproducible for the semiquantitation of SCFAs. Stability was studied using the same sample at 0, 1, 2, 4 and 6 h. RSDs were found to be below 6.39%, which indicated that the sample solution was stable in an ice box until 6 h.
In order to further ensure the reliable and reproducible nature of the determination results throughout the entire study, a sample from the method validation was set as the quality control (QC) sample to monitor sample determination. Specifically, the QC sample was analyzed at intervals of five batches of samples.

Characteristic chromatographic fingerprint study and PCA
The GC-MS method was subsequently applied to analyze SCFAs in 70 milk samples, including 45 samples  of human milk, 12 samples of pure milk, eight samples of fermented milk and five samples of infant formula. The representative SIM chromatograms of the sample solutions were shown in Figure 1. The semi-quantitative results of the human milk and other milk products (infant formula, pure milk and fermented milk) were summarized in Tables 3 and 4, respectively. The relative concentration trends of SCFAs in all samples were visually exhibited by heat map in Figure 2. The semi-quantitative results were calculated according to the following equations: R: response, the semi-quantitative results; A S : integral area of SCFAs in samples; A IS : integral area of IS with a fixed concentration.
Accordingly, the total quantity of SCFAs in human milk, infant formula, pure milk and fermented milk was 17.55, 7.11, 8.96 and 25.13, respectively. Moreover, the major SCFAs in human milk were shown to be acetic acid (38.23%), formic acid (18.01%), butyric acid (16.24%) and hexanoic acid (25.59%). Acetic acid and formic acid were the predominant SCFAs in dairy products: infant formula contained 17.54% acetic acid and 75.71% formic acid, pure milk had 53.36% and 28.34%, respectively, and fermented milk contained 43.64% and 45.57%, respectively. The data distribution (median, minimum, maximum, and the first and third quartiles) of SCFAs in human milk and other milk products were shown in Figure S3. In conclusion, no matter the quantity or proportion of SCFAs, milk products were significantly different compared to human milk.
For further analysis, an unsupervised multivariate technique, PCA, was used to explore the relationships between our observations and any variables. Both score and loading plots for the PCA were constructed based on the first two principal components (PCs), which accounted for 89.16% of the total variability (Figure 3). PC1 explained the 48.96% of the total variation and PC2 explained 40.20%. As shown in Figure 3(A), human milk, infant formula, pure milk and fermented milk were divided into four distinct groups, which  In the score plot, human milk was separated from pure milk, indicating that the composition of SCFAs was significantly different. The result was consistent with the concept expressed by Ley et al. (2008), which could be explained primarily by the difference in gut microbial composition between humans and cows, owing to both diet and phylogeny. In addition, 12 batches of pure milk, including local and imported brands, were close to each other, whereas 45 samples of human milk in a group were markedly scattered. Dietary habits were considered to be one of the major contributory factors to this finding (B€ ackhed et al. 2005). Similar feeding patterns and forage promote the comparatively unanimous diversity of cow gut microbiota. Instead, the variety of food, dietary habits and lifestyles result in a diverse composition of gut microbiota in humans.
All batches of infant formula were clustered at the bottom of the score plot. This result could be explained by a relatively low quantity of SCFAs in the formula samples, which may be related to the drying process of infant formula that would remove a proportion of the SCFAs along with the evaporation of water.
Fermented milk samples were scattered across the first quadrant of the score plot, distant from the other groups. This separation was a result of differences in the proportion of SCFAs among fermented milk and other milk samples; fermentation occurred as a consequence of certain specific lactic acid-producing bacteria (Sachdeva et al. 2014), that differ dramatically from cow and human gut microbiota. Furthermore, the duration and specifics of the production processes are major variables (Sachdeva et al. 2014) that can introduce variations into the quantity and species of SCFAs that relate to flavor and quality (Routray & Mishra 2011). The bacteria found in the fermented milk samples used in this study are summarized in Table S3, showing the influence of bacteria on the fermentation process.
The loading plot was performed to evaluate the influence of the original variables associated with the clustering of samples and explore the most important variables. As shown in Figure 3(B), acetic acid, formic acid, butyric acid and hexanoic acid remarkably deviated from the origin of the loading plot. These factors could be regarded as chemical markers to discriminate human milk from infant formula, pure milk and fermented milk. Furthermore, benefiting from the ideal composition of nutrients and comprehensive functions of SCFAs (shown in Figure 4), the SCFAs in human milk could be considered as potential reference for infant formula and other milk products.

Conclusion
This work has established a fingerprinting method based on a GC-MS technique for the effective discrimination of different milk types and dairy products using chemometric techniques. A simple sample preparation step followed by GC-MS detection was developed to investigate SCFAs in human milk, infant formula, pure milk and fermented milk. Definitive distinctions between the milk types were achieved easily using PCA. The results indicated that the content of SCFAs in human milk, infant formula and dairy products is significantly different. This study provides a valuable  reference for the further development of protocols that assess the chemical composition of human milk and various dairy products.