Abstract

   Organic milk has a high risk of food fraud as it can easily be
   adulterated with non-organic milk. This study aimed to identify
   metabolite markers for assessing the authenticity of organic milk from
   Jersey and Yak. In the untargeted strategy, ultra-high performance
   liquid chromatography-Q Exactive HF-X mass spectrometer coupled with
   chemometrics analysis was used to screen and identify tentative markers
   of organic milk from Jersey and Yak. In the targeted strategy, a quick
   and easy method of ultra-performance liquid chromatography–tandem mass
   spectrometry (UPLC-MS/MS) was developed to quantify three markers. The
   peptide of Thr-Ala-Val and D-biotin were determined to be metabolite
   markers for distinguishing organic and non-organic Jersey milk, whereas
   trimethylamine N-oxide was determined to be a metabolite marker for
   distinguishing organic and non-organic Yak milk. These findings provide
   critical information to facilitate assessments of organic milk
   authenticity.

1. Introduction

   Bovine milk, as a natural food with high nutritional value, is favored
   and strongly recommended as a high-quality protein source in dietary
   guidelines ([35]Kang et al., 2022a, [36]Nagpal et al., 2012). According
   to the type of production management system, bovine milk can be divided
   into organic and traditional milk. Compared to traditional milk,
   organic milk contains more omega-3 polyunsaturated fatty acids, which
   are important for the prevention of certain types of malignant diseases
   and autoimmune disorders ([37]Mie et al., 2017, [38]Mariamenatu and
   Abdu, 2021). Organic milk is a highly safe product because the use of
   synthetic chemical substances, such as chemical fertilizers,
   pesticides, hormones, growth regulators, feed additives, and food
   additives, is strictly prohibited in production and processing. In
   recent years, sales of organic milk products have increased
   significantly in the market and are up to 50% more expensive than
   traditional milk ([39]Chung et al., 2018). Driven by economic profits,
   some traders adulterate organic milk with relatively cheap non-organic
   milk, and this adulteration phenomenon reduces the product quality and
   damages the rights of consumers. In China, cow breeds producing organic
   milk mainly include Holstein, Jersey, and Yak, of which Jersey and Yak
   are minor dairy animals. Both Jersey and Yak milk are known as premium
   milk, with lower production levels and higher prices than that of
   Holstein milk. Moreover, the knowledge of Jersey and Yak organic milk
   is poor. Therefore, it is urgently required to protect the consumer
   from wrongly labeled Jersey and Yak organic milk.

   To date, authenticity identification technologies for organic milk have
   mainly focused on stable isotope analysis. Carbon and nitrogen isotope
   ratios (δ^13C and δ^15N) in milk can vary considerably depending on the
   cow's diet ([40]Molkentin, 2013). The structure of pasture forage
   directly affects the ratio of C3 to C4 from the photosynthetic plants
   consumed in the dairy cow diet, resulting in differences of δ^13C value
   between organic and non-organic milk ([41]Chung et al., 2014;
   [42]Molkentin, 2013). Compared with manure allowed by organic farms,
   the δ^15N value of chemical synthetic nitrogen fertilizers is lower.
   Because organic farms do not use chemical synthetic fertilizers, the
   δ^15N value of organic milk was significantly different from
   conventional milk ([43]Chung et al., 2014). Therefore, the δ^13C and
   δ^15N values were applied to distinguish organic and non-organic milk
   ([44]Chung et al., 2018, [45]Chung et al., 2014; [46]Molkentin, 2009).
   However, the isotope ratio mass spectrometer is expensive, and the
   technology is difficult to translate by conventional instruments, which
   makes this technology difficult to use in actual detection.
   Metabolomics have attracted increasing attention because of their high
   selectivity for characteristic markers and high accuracy in detecting
   adulteration. According to recently published data, metabolomics have
   been confirmed for the identification of milk from different breeds of
   animals and different feeding systems based on untargeted metabolomics
   ([47]Caboni et al., 2019; [48]Ji et al., 2023, [49]Fan et al., 2023;
   [50]Qin et al., 2022, [51]Rocchetti et al., 2022, [52]Scano et al.,
   2014, [53]Sen et al., 2021; [54]Yang et al., 2016). However, to date,
   there is a lack of studies about the use of metabolomic methods to
   distinguish organic and non-organic milk, especially Jersey and Yak
   milk. Thus, we aimed to establish an analytical strategy to distinguish
   organic and non-organic milk (both Jersey and Yak milk) based on
   metabolic markers, which could fill the gap in the identification of
   Jersey and Yak organic milk.

   In this study, we established untargeted and targeted metabolomics
   methods to reveal the authenticity of Jersey and Yak organic milk based
   on liquid chromatography-mass spectrometry (LC-MS). Specifically,
   high-resolution mass spectrometry combined with chemometrics was used
   to screen potential markers which are used to distinguish organic and
   non-organic milk (both Jersey and Yak milk), and a quantitative
   approach was developed by ultra-performance liquid
   chromatography-tandem mass spectrometry (UPLC-MS/MS) to verify the
   potential markers and determine content thresholds of markers.

2. Materials and methods

2.1. Chemicals and solvents

   HPLC-grade methanol, acetonitrile, and water were purchased from Fisher
   Scientific (Pittsburgh, PA), HPLC-grade isopropanol was purchased from
   Merck (Darmstadt, Germany), and HPLC-grade formic acid was purchased
   from CNW (Shanghai, China). Trimethylamine N-oxide was obtained from
   Yuanye Bio-technology (Shanghai, China). Biotin was purchased from
   Solarbio (Beijing, China). Peptide of Thr-Ala-Val was synthesized by TG
   peptide (Nanjing, China). L-2-chlorophenylalanin was obtained from
   Adamas-beta (Shanghai, China). All standards had a purity of ≥ 95%.

2.2. Samples information

   Twenty samples of each variety of milk were collected during July and
   September 2022, including organic Yak milk (OYM), non-organic Yak milk
   (NOYM), organic Jersey milk (OJM), and non-organic Jersey milk (NOJM).
   The OYM samples were collected from Lhasa city (Tibet, China). NOYM
   samples were collected from Lhasa city (Tibet, China) and Qilian County
   (Qinghai, China). OJM samples were collected from Hohhot (Inner
   Mongolia, China). NOJM samples were obtained from Hangzhou city
   (Zhejiang, China) and Hohhot (Inner Mongolia, China). After ultra-high
   temperature instant sterilization, all samples were sealed and
   packaged. Additionally, four different adulteration milk ratios were
   prepared by mixing 10, 20, 30, and 60% non-organic milk in organic milk
   (both Yak and Jersey milk). Five adulterated simulants were prepared
   for each type in Yak and Jersey milk. In total, 80 real samples and 40
   adulterated simulants were finally obtained. The liquid samples were
   vacuum freeze-dried to a loose dry state.

2.3. Untargeted metabolomic analysis

2.3.1. Metabolite extraction

   20 mg of milk powder, a 6 mm diameter grinding bead, and 400 µL
   methanol–water (4:1, v/v) solution with 0.02 mg/mL internal standard of
   L-2-chlorophenylalanin were added sequentially to a 2 mL centrifuge
   tube. The mixture was allowed to settle at −10 °C and ground for 6 min,
   then followed by treatment with 40 kHz ultrasound at 5 °C for 30 min.
   The extraction solution was placed at −20 °C for 30 min and centrifuged
   for 15 min (13,000 g, 4 °C). After the supernatant was dried with
   nitrogen, the obtained extracts were redissolved by 120 µL
   acetonitrile–water (1:1, v/v) solution. The solution was vortexed for
   30 s and sonicated for 5 min (5 °C, 40 kHz). After centrifugation for
   15 min (13,000 g, 4 °C), the supernatant was collected for LC-MS
   analysis. In addition, 20 µL of supernatant was removed from each
   sample and mixed as a quality control (QC) sample.

2.3.2. Instrumentation

   For untargeted metabolomics analysis, 3 µL of the sample was separated
   and analyzed by using UHPLC-Q Exactive HF-X Mass Spectrometer of Thermo
   Fisher Scientific equipped with an ACQUITY HSS T3 column
   (100 mm × 2.1 mm i.d. × 1.8 μm; Waters, Milford, USA). The mobile
   phases consisted of solvent A (0.1% formic acid in water:
   acetonitrile = 95:5, v/v) and solvent B (0.1% formic acid in
   acetonitrile: isopropanol: water = 47.5:47.5:5, v/v). The gradient used
   was: 0%-20% (B), 0–3 min; 20%-35% (B), 3–4.5 min; 35%-100% (B),
   4.5–5 min; 100% (B), 5–6.3 min; 0% (B), 6.3–8 min. The flow rate was
   0.4 mL/min, and the column temperature was 40 °C.

   The mass spectrometric data was collected using a UHPLC -Q Exactive
   HF-X Mass Spectrometer equipped with an electrospray ionization (ESI)
   source operating in either positive or negative ion mode. The optimal
   conditions were: sheath gas flow rate, 50 arb; aux gas flow rate, 13
   arb; heater temperature, 425 °C; capillary temperature, 325 °C;
   normalized collision energy, 20–40-60 V rolling for MS/MS; the spray
   voltage was set at (+) 3500 V and (−) 3500 V, respectively; full MS
   resolution was 60,000, and MS/MS resolution was 7500; the range of the
   MS scan was from m/z 70 to 1050. Moreover, a quality control (QC)
   sample was randomly inserted into every 3 analysis samples to evaluate
   the repeatability and robustness of the analysis.

2.3.3. Data preprocessing and annotation

   Raw data of LC-MS was imported into the Progenesis QI software (Waters,
   Milford, USA) for baseline filtering, peak identification, integration,
   retention time correction, and peak alignment. Subsequently, the MS and
   MS/MS mass spectra were matched with the metabolite database (the main
   databases were the HMDB ([55]https://www.hmdb.ca/), Metlin
   ([56]https://metlin.scripps.edu/), and Majorbio Database). The MS mass
   error was set to at least 10 ppm, while the metabolites were identified
   based on the secondary mass spectra matching score. Metabolites
   detected at least 80% in any set of samples were retained. At the same
   time, variables with relative standard deviation (RSD) > 30% of QC
   samples were discarded and response intensity of the mass spectrum
   peaks were log-transformed to obtain the final data matrix for
   subsequent analysis.

2.3.4. Differential metabolites analysis

   Statistical analyzes were performed with R and Python packages, such as
   hierarchical cluster analysis, principal component analysis (PCA),
   least partial squares discriminant analysis (PLS-DA), student's t-test,
   and fold difference analysis. Differential metabolites were selected
   based on the variable importance in the projection (VIP) obtained by
   the PLS-DA model, the p-value of the student’s t test, and fold change
   (FC). Then, these differential metabolites were summarized through
   metabolic enrichment and pathway analysis based on database search
   (KEGG, [57]https://www. [58]genome.jp/kegg/).

2.4. Targeted metabolomic analysis

2.4.1. Metabolite extraction

   200.0 mg samples were placed into a 10 mL centrifuge tube containing
   1.0 mL of water, vortexed until the milk powder was completely
   dissolved. After adding 4 mL of methanol/acetonitrile (1:1, v/v), the
   mixture was vortexed for 30 s, then followed by treated with 40 kHz
   ultrasound at 5 °C for 40 min. The extraction solution was placed at
   −20 °C for 30 min and centrifuged for 20 min (8500 rpm, 4 °C). Finally,
   the supernatant was collected for UPLC-MS/MS analysis.

2.4.2. Instrumentation

   Quantification of the selected substances was performed on Agilent 1200
   HPLC-6420 Triple Quad LC/MS (Agilent Technologies) equipped with a
   degasser, binary pump, auto-sampler, and column oven. Separation of the
   analytes was performed using an Agilent ZORBAX SB-C18 column
   (150 × 4.6 mm, 5 µm) at 40 °C. The mobile phases consisted of solvent A
   (0.1% formic acid in water: acetonitrile = 95:5, v/v) and solvent B
   (0.1% formic acid in acetonitrile: isopropanol: water = 47.5:47.5:5,
   v/v). The gradient used was: 5%-10% (B), 0–3 min; 10% (B), 3–4 min;
   10%-50% (B), 4–5 min; 50%-100% (B), 5–6 min; 100% (B), 6–7.5 min;
   100%-35% (B), 7.5–9 min; 35%-5% (B), 9–10 min; 5% (B), 10–12 min. The
   injection volume was 5 µL.

   Detection was carried out in the positive ion mode with a capillary
   voltage of 4.0 kV. The other instrumental parameters were set as
   follows: drying gas temperature, 350 °C; drying gas flow, 10 L/min; and
   nebulizer, 40 psi. Multiple Reaction Monitoring (MRM) parameters for
   the three target analytes were summarized in [59]Table 1.

Table 1.

   Molecular weight and optimized MS parameters for the three potential
   markers in the ESI + mode.
   Compound Retention time (min) Precursor ions (m/z) Product ions (m/z)
   Fragmentor (V) Collision energy (eV)
   Trimethylamine N-oxide 3.29 76.1 42.3 35 53
   58.3* 21
   59.3 9
   Thr-Ala-Val 3.74 290.3 101.1 25 83
   145 13
   173* 9
   Biotin 9.25 245.3 97.1 33 93
   123 33
   227.1* 13
   [60]Open in a new tab

   * Quantification ion.

2.4.3. Statistical analysis

   Statistical analyzes were performed with R packages, while one-way
   analysis of variance (ANOVA) and box diagram were employed to confirm
   significant differences and content thresholds of the final markers in
   organic and non-organic milk with p < 0.05.

3. Results and discussion

3.1. Non-targeted metabolic profiling of four cultivars of cow milk

   We used the internal standard of L-2-chlorophenylalanin to assess the
   stability of the experimental procedure and ensure data quality. The
   z-score values of the internal standard were within two times the
   standard deviation, which indicated the stable data. Therefore, we did
   not normalize the data (e.g., sum normalization) and directly used the
   ion intensity for the subsequent statistical analysis. The QC samples
   were evaluated to ensure the repeatability and robustness of the whole
   sample set ([61]Broadhurst et al., 2018): Total ion chromatograms
   (TICs) of the QC sample in the positive and negative ion scanning modes
   were shown in [62]Fig. S1, the peak shape and distribution are
   relatively uniform; the QC samples showed a tight cluster in the PCA
   plot ([63]Fig. S1). The above assessment results for the QC samples
   demonstrated the robustness of the analytical procedure and the
   reliability of the data obtained. Totally, 7465 and 2582 features were
   detected, respectively, in the positive and negative ion mode, of which
   730 and 305 metabolites were annotated, respectively. Since the
   metabolites identified in the positive ion model are more abundant, we
   selected the metabolite profiles of the positive ion mode for
   subsequent analyses.

   To characterize the similarity between four cultivars of cow milk,
   hierarchical cluster analysis of the metabolite ion features was
   carried out and represented in the form of a heatmap. The color coding
   from red to blue indicates their content from high to low. In [64]Fig.
   1, a significant color distribution is observed, indicating that
   different milk samples presented differential metabolite profiles. Four
   milks were clearly clustered into two different sections, revealing
   significant differences between Yak and Jersey milk. Not surprisingly,
   organic and non-organic milk from the same breed of cow were clustered
   into one group for both Jersey and Yak, which meant similarity of their
   metabolite profiles. However, the significant color distribution
   between organic and non-organic milk also revealed their differences.
   It is very meaningful to characterize the differential metabolites
   between organic and non-organic milk to prevent food fraud.

Fig. 1.

   Fig. 1
   [65]Open in a new tab

   Heatmap of cluster analysis for metabolite profiles in four cultivars
   of cow milk (organic Yak milk (OYM), non-organic Yak milk (NOYM),
   organic Jersey milk (OJM), and non-organic Jersey milk (NOJM)).

3.2. Metabolic characteristics of organic Jersey milk

   To dissect the differences in metabolite composition between organic
   and non-organic Jersey milk (OJM and NOJM), unsupervised PCA and
   supervised PLS-DA were carried out. The results of PCA were shown in
   [66]Fig. 2a, OJM and NOJM samples exhibited a distinct separation. The
   clear clustering indicated that there was a significant difference
   between the OJM and NOJM. The first two PCs showed 29.8 % and 24.9 %
   variances, respectively. The results of PLS-DA were similar to PCA
   ([67]Fig. 2b), the well-separated clusters of OJM and NOJM samples
   exhibited that dairy farming management systems had a strong influence
   on the metabolite profile of Jersey milk. The quality of PLS-DA model
   developed was confirmed by a good R^2Y of 1.0 and Q^2 of 0.90
   ([68]Wang, Li, Chen, & Zhou, 2022). The variable importance in
   projection (VIP) analysis based on PLS-DA model was applied to identify
   the important metabolites for distinguishing OJM and NOJM groups
   ([69]Kang et al., 2022b), and there were 1517 metabolites with VIP
   scores > 1.

Fig. 2.

   [70]Fig. 2
   [71]Open in a new tab

   Multivariate statistical analysis of metabolites in Jersey milk. a. PCA
   of metabolites from the OJM and NOJM groups; b. supervised PLS-DA of
   metabolites from the OJM and NOJM groups; c. volcano plots of
   differential metabolites between OJM and NOJM groups; d. metabolite
   pathway analysis on differential metabolites from the pairwise
   comparison “OJM vs. NOJM”.

   To obtain differential metabolites of the OJM and NOJM samples, the
   p-value and fold change (FC) of univariate analysis were further
   performed on these 1517 metabolites ([72]Wang et al., 2022). Selection
   criteria included VIP scores > 1, p-value < 0.05, and FC > 1 or < 1,
   and the results were visualized using a volcano plot. Finally, a total
   of 509 metabolites with significant differences were screened and 85 of
   them were further annotated ([73]Fig. 2c and [74]Table 2), containing
   primarily 23 amino acids, peptides, and analogues, 3 fatty acids and
   conjugates, 3 indoles, 2 triterpenoids, 2 glycerophosphocholines, and 2
   diterpenoids. It was concluded that amino acids, peptides, and
   analogues are the largest different chemicals. These 85 differential
   metabolites were selected as the potential markers to distinguish NOJM
   from OJM. In [75]Fig. 2c, the red dots represent 51 differential
   metabolites with higher content in OJM and the blue dots represent 34
   differential metabolites with higher content in NOJM. Interestingly,
   the levels of seven most up-regulated metabolites were above 28 times
   higher than those in NOJM, namely domoic acid, notoginsenoside C,
   Lys-Ile-Glu, isoputreanine,
   5,6-dihydro-5-hydroxy-6-methyl-2H-pyran-2-one, Arg-Leu, and
   16-oxoestrone. Notoginsenoside C is a bioactive saponins from the root
   of panax notoginseng ([76]Yoshikawa et al., 1997); the
   5,6-dihydro-5-hydroxy-6-methyl-2H-pyran-2-one was a natural product
   from fern species ([77]Gyeltshen et al., 2022); their high content in
   OJM reflects the characteristics of organic pasture. Most peptides in
   differential metabolites exhibited up-regulation in OJM, such as
   kyotorphin, Arg-Asn, Ile-Ile-Asn, Ile-Ile-Val, Val-Leu-Pro-Val-Pro,
   Thr-Ala-Val, Lys-Ile-Glu, and Arg-Leu. Previous studies have shown that
   kyotorphin, first isolated from bovine brain in 1979, is an endogenous
   analgesic neuropeptide with anti-inflammatory and antibacterial
   activity ([78]De Andrade et al., 2020); Val-Leu-Pro-Val-Pro is an
   angiotensin converting enzyme (ACE) inhibitory peptide that plays a
   very important role in blood pressure regulation ([79]Lei et al.,
   2008). These results provided some basis for confirming that OJM has a
   better nutritional value than NOJM.

Table 2.

   The information of all potential differential metabolites from the
   pairwise comparison “OJM vs. NOJM”.
   ID Metabolite Formula M/Z Adducts Regulate VIP FC p-value
   1 LysoPC (18:3(6Z,9Z,12Z)/0:0) C26H48NO7P 518.3224 M + H, M + Na up
   1.59 1.09 0.01
   2 Cholesteryl glucoside C33H56O6 566.4391 M + NH4, M + Na up 1.08 1.04
   0.03
   3 l-Serine C3H7NO3 147.076 M + H-H2O, M + ACN + H up 1.15 1.05 0.00
   4 Isoputreanine C7H16N2O2 161.1279 M + H up 4.56 28.76 0.00
   5 Nitrilotriacetic acid C6H9NO6 424.1198 2 M + ACN + H up 1.09 1.05
   0.02
   6 Triethylcitrate C12H20O7 299.1091 M + H, M + NH4, M + Na down 1.08
   0.96 0.02
   7 Aminomalonic acid C3H5NO4 152.0562 M + CH3OH + H up 1.01 1.03 0.01
   8 Triisopropanolamine C9H21NO3 192.1589 M + H down 2.28 0.82 0.00
   9 Histidylleucine C12H20N4O3 269.1599 M + H up 1.55 1.11 0.00
   10 N-Acetyl-l-Histidine C8H11N3O3 198.0871 M + H up 2.05 1.20 0.00
   11 6-Sulfanilamidoindazole C13H12N4O2S 253.0535 M + H-2H2O up 4.17 4.34
   0.03
   12 5,6-Dihydro-5-hydroxy-6-methyl-2H-pyran-2-one C14H12O3 292.0951
   M + ACN + Na up 4.66 28.76 0.00
   13 1-[2-Methyl-3-(methylsulfanyl)propanoyl]pyrroli-dine-2-carboxylate
   C10H17NO3S 276.0639 M + 2Na-H up 1.09 1.04 0.00
   14 Thr-Ala-Val C12H23N3O5 290.1699 M + H up 2.64 1.27 0.00
   15 4-Hydroxy-4-(3-pyridyl)-butanoic acid C9H11NO3 199.107 M + NH4 down
   1.07 0.96 0.01
   16 Kyotorphin C15H23N5O4 338.1808 M + H up 2.17 1.19 0.01
   17 Tetrahydroharmol C12H14N2O 203.1173 M + H down 2.63 0.79 0.00
   18 Domoic acid C15H21NO6 334.1242 M + Na up 4.37 28.76 0.00
   19 2,3-Methyleneglutaric acid C7H8O4 121.0283 M + H-2H2O up 1.11 1.06
   0.01
   20 Alpha-Methyltryptamine C11H14N2 175.1225 M + H up 1.11 1.05 0.00
   21 5-Methylthioribose C6H12O4S 222.0789 M + ACN + H down 4.34 0.19 0.01
   22 Arg-Asn C10H20N6O4 577.3176 2 M + H up 2.71 1.28 0.00
   23 LysoPS (18:0/0:0) C24H48NO9P 526.3123 M + H, M + Na, M + 2Na-H up
   1.52 1.07 0.01
   24 S-4-Hydroxymephenytoin C12H14N2O3 199.086 M + H-2H2O down 2.08 0.87
   0.00
   25 5-Hydroxy-l-tryptophan C11H12N2O3 243.0758 M + Na up 1.44 1.07 0.00
   26 Indoleacetic acid C10H9NO2 217.0966 M + ACN + H down 2.11 0.86 0.00
   27 Nalidixic acid C12H12N2O3 215.081 M + H-H2O down 1.89 0.90 0.00
   28 D-Biotin C10H16N2O3S 245.0947 M + H down 2.49 0.80 0.00
   29 Tryptophol C10H11NO 144.0804 M + H-H2O down 1.82 0.86 0.02
   30 Butyl 4-aminobenzoate C11H15NO2 158.096 M + H-2H2O down 1.80 0.90
   0.02
   31 Alpha-Linolenoyl ethanolamide C20H35NO2 344.2532 M + Na up 1.46 1.06
   0.00
   32 10-Deacetylbaccatin III C29H36O10 562.2679 M + NH4 up 1.50 1.10 0.04
   33 Undecylenic acid C11H20O2 202.1795 M + NH4 down 2.59 0.64 0.05
   34 Panaxydol linoleate C35H54O3 540.4453 M + NH4 up 1.33 1.06 0.05
   35 Glucosyl (2E,6E,10x)-10,11-dihydroxy-2,6-farnesadienoate C21H36O9
   415.235 M + H-H2O up 1.93 1.19 0.01
   36 Lupeol acetate C32H52O2 507.364 M + K down 1.18 0.96 0.00
   37 Polypodine B C27H44O8 535.2686 M + K up 1.10 1.04 0.01
   38 Cer (d17:1/PGJ2) C37H63NO5 584.4709 M + H-H2O up 1.55 1.09 0.05
   39 Ganoderic acid F C32H42O9 609.2479 M + K up 1.05 1.04 0.01
   40 16-Oxoestrone C18H20O3 569.2914 2 M + H up 3.46 28.76 0.00
   41 Ile-Ile-Val C17H33N3O4 344.2532 M + H, M + K, 2 M + H, M + Na up
   1.58 1.09 0.01
   42 1-(7Z,10Z,13Z,16Z-docosatetraenoyl)-glycero-3-phosphate C25H43O7P
   528.3067 M + ACN + H up 1.98 1.19 0.03
   43 Cer (d18:2(4E,14Z)/20:5 (7Z,9Z,11E,13E,17Z)–3OH (5,6,15)) C38H63NO6
   630.4767 M + H down 1.06 0.97 0.00
   44 KAPA C9H17NO3 188.1276 M + H, M + Na, M + H-H2O, M + H-2H2O up 1.22
   1.05 0.03
   45 Ser-Phe C12H16N2O4 217.0968 M + H-2H2O, M + K down 2.20 0.84 0.00
   46 PC (10:0/0:0) C18H38NO7P 412.2443 M + H down 1.02 0.96 0.03
   47 Xanthosine C10H12N4O6 267.0731 M + H-H2O up 1.19 1.06 0.00
   48 PGP (PGE2/18:3 (9Z,12Z,15Z)) C44H74O16P2 472.2226 M + H + Na,
   M + 2Na up 1.07 1.04 0.04
   49 Cyclo (aspartylleucylthreonylvalyltyrosylphenylalanylglycyl)
   C39H53N7O11 809.3983 2 M + 3H2O + 2H up 1.08 1.04 0.03
   50 Val-Leu-Pro-Val-Pro C31H53N7O8 652.4009 M + 2H, M + H up 1.89 1.17
   0.02
   51 Diphenylamine C12H11N 187.1223 M + H, M + NH4 down 1.95 0.90 0.00
   52 2-(4-(Piperidin-3-yl) phenyl)–2H-indazole-7-carboxamide C19H20N4O
   663.3153 2 M + Na up 1.99 1.16 0.00
   53 (R)-beta-Aminoisobutyric acid C4H9NO2 245.0914 2 M + K up 1.40 1.07
   0.01
   54 N-Acetyl-d-tryptophan C13H14N2O3 247.107 M + H up 1.38 1.07 0.02
   55 3-Methyl-N-phenylaniline C13H13N 201.138 M + H, M + NH4 down 2.33
   0.82 0.00
   56 Notoginsenoside C C54H92O25 571.3022 M + 2H up 4.95 28.76 0.00
   57 Methyl 4-amino-5-ethyl-3-thiophenecarboxylate C8H11NO2S 186.0577
   M + H down 1.24 0.94 0.00
   58 Quinaldic acid C10H7NO2 215.081 M + ACN + H down 1.95 0.88 0.00
   59 Carbazole C12H9N 185.1068 M + NH4 down 1.88 0.88 0.00
   60 5-Hydroxyindoleacetaldehyde C10H9NO2 217.0967 M + ACN + H down 3.27
   0.70 0.00
   61 1,5-Naphthalene diisocyanate C12H6N2O2 243.0758 M + CH3OH + H up
   1.32 1.07 0.01
   62 Taxine B C33H45NO8 584.3174 M + H up 2.13 1.20 0.01
   63 Flavin Mononucleotide C17H21N4O9P 457.1105 M + H down 1.32 0.94 0.00
   64 Riboflavin cyclic-4′,5′-phosphate C17H19N4O8P 439.1 M + H up 1.14
   1.05 0.03
   65 Fusaric Acid C10H13NO2 197.1279 M + NH4 up 2.21 1.18 0.00
   66 Ile-Ile-Asn C16H30N4O5 359.2277 M + H up 1.18 1.06 0.03
   67 Harmalol C12H12N2O 201.1017 M + H down 2.60 0.78 0.00
   68 1,2,5,6-Tetrahydro-4H-pyrrolo[3,2,1-ij]quinolin-4-one C11H11NO
   174.0909 M + H down 2.62 0.77 0.00
   69 3-Indolebutyric acid C12H13NO2 186.0908 M + H-H2O down 2.22 0.82
   0.00
   70 Methylene bisacrylamide C7H10N2O2 326.181 2 M + NH4 up 1.38 1.07
   0.00
   71 2-(1-Naphthyl) acetamide C12H11NO 203.1172 M + NH4 down 2.18 0.81
   0.00
   72 Norvaline C5H11NO2 100.0758 M + H-H2O down 1.12 0.95 0.02
   73 4-Methyl-5-Thiazoleethanol C6H9NOS 144.0474 M + H up 1.12 1.06 0.05
   74 Tetrahydrouridine C9H16N2O6 271.0914 M + Na down 1.11 0.94 0.01
   75 2-Isopropyl-3-oxosuccinate C7H10O5 157.0493 M + H-H2O down 1.78 0.85
   0.01
   76 Butenylcarnitine C11H19NO4 230.138 M + H up 1.23 1.06 0.00
   77 Arg-Leu C12H25N5O3 288.2019 M + H up 4.69 28.76 0.00
   78 5′-Deoxyadenosine C10H13N5O3 252.1083 M + H up 3.51 1.61 0.00
   79 Lys-Ile-Glu C17H32N4O6 389.238 M + H up 4.50 28.76 0.00
   80 D-4′-Phosphopantothenate C9H18NO8P 300.0832 M + H down 1.09 0.96
   0.01
   81 4-Guanidinobutanoic Acid C5H11N3O2 146.092 M + H up 1.51 1.10 0.02
   82 Mimosine C8H10N2O4 216.0973 M + NH4 up 1.98 1.17 0.01
   83 Prenyl arabinosyl-(1->6)-glucoside C16H28O10 403.1545 M + Na down
   1.12 0.96 0.01
   84 Calcium pantothenate C18H32CaN2O10 477.174 M + H, M + Na down 1.01
   0.96 0.01
   85 Methyl methacrylate C5H8O2 218.138 2 M + NH4 down 1.22 0.94 0.00
   [80]Open in a new tab

   In addition, a pathway enrichment analysis was performed on these
   differential metabolites to explain the relationship between variation
   in the chemical composition and dairy farming management systems. The
   enrichment background in KEGG annotation was all metabolites of dairy
   and the results were represented by a bubble plot ([81]Fig. 2d), in
   which the abscissa is rich factor (x/y, the number of differential
   metabolites in the corresponding metabolic pathway / the number of
   total metabolites in this pathway); the color and size of the bubble
   indicate the p-value and the number of differential metabolites. We
   found that differential metabolites between OJM and NOJM were assigned
   in 31 metabolic pathways, of which eight main pathways with significant
   enrichment were determined (p < 0.05): biosynthesis of cofactors,
   tryptophan metabolism, cysteine and methionine metabolism, ABC
   transporters, glycerophospholipid metabolism, vitamin digestion and
   absorption, biotin metabolism, and choline metabolism in cancer. These
   eight pathways contributed to the metabolic differences between OJM and
   NOJM. It was also concluded that the organic feeding management system
   was found to possess a high impact on these metabolic pathways of
   Jersey. It is noteworthy that tryptophan metabolism was greatly
   affected and four differential metabolites, including tryptophol,
   indoleacetic acid, 5-hydroxyindoleacetaldehyde, and
   5-hydroxy-l-tryptophan, were enriched. Amino acid metabolism pathways
   composed of valine, leucine, isoleucine, glycine, serine, threonine,
   arginine, proline, cysteine, methionine, and tryptophan may be a
   possible biological process for the different chemicals relating to the
   forage structure. Similarly, a previous study has shown that the
   arginine biosynthesis and some amino acid metabolisms of dairy cow are
   significantly affected by different feeding strategies (based on corn
   silage as the main ingredient) ([82]Rocchetti et al., 2022). As
   mentioned above, these results could provide new insights into the
   effects of organic production systems on metabolism of Jersey.

3.3. Metabolic characteristics of organic Yak milk

   To dissect the differences in metabolite composition between organic
   and non-organic Yak milk (OYM and NOYM), unsupervised PCA and
   supervised PLS-DA were carried out. The results of PCA were shown in
   [83]Fig. 3a, OYM and NOYM samples exhibited a large degree of
   overlapping, which was indicative of their similarity. But the two
   groups had a tendency to separate, and the first two PCs showed 33.8 %
   and 21.4 % variances, respectively. To further determine the
   metabolites to distinguish NOYM from NOYM samples, the PLS-DA model was
   established ([84]Fig. 3b), which showed well repeatability of samples
   within the same group and significant separation degree between OYM and
   NOYM. The quality of PLS-DA model developed was confirmed by a good
   R^2Y of 0.996 and Q^2 of 0.918 ([85]Shi et al., 2022). The VIP analysis
   based on PLS-DA model was applied to identify the important metabolites
   for distinguishing OYM and NOYM groups ([86]Kang et al., 2022a), and
   there were 1537 metabolites with VIP scores > 1.

Fig. 3.

   [87]Fig. 3
   [88]Open in a new tab

   Multivariate statistical analysis of metabolites in Yak milk. a. PCA of
   metabolites from the OYM and NOYM groups; b. supervised PLS-DA of
   metabolites from the OYM and NOYM groups; c. volcano plots of
   differential metabolites between OYM and NOYM groups; d. metabolite
   pathway analysis on differential metabolites from the pairwise
   comparison “OYM vs. NOYM”.

   To obtain the differential metabolites of OYM and NOYM samples, the
   p-value and FC of univariate analysis were further performed on these
   1537 metabolites. Selection criteria included VIP scores > 1,
   p-value < 0.05, and FC > 1 or < 1, and the results were visualized
   using a volcano plot. Finally, a total of 156 metabolites with
   significant differences were screened and 34 of them were annotated
   ([89]Fig. 3c and [90]Table 3), mainly including 8 amino acids,
   peptides, and analogues, 3 carbohydrates and carbohydrate conjugates,
   and 2 carbonyl compounds. It was concluded that amino acids, peptides,
   and analogues were the largest different chemicals between OYM and
   NOYM, which is similar to Jersey milk. In [91]Fig. 3c, the red dots
   represent 6 differential metabolites with higher content in OYM, and
   the blue dots represent 28 differential metabolites with higher content
   in NOYM. The top 2 most up-regulated metabolites in OYM were
   2′-deoxyadenosine 5′-phosphate and hydroxykynurenine, their FC values
   are 5.29 and 2.71, respectively. Interestingly, the FC values of
   omphalotin B and PS (22:2(13Z,16Z)/TXB2) are 0 because their response
   intensity in OYM is lower than the instrument detection limit.
   Omphalotin B, a bioactive ingredient, is a nematicidal cyclic
   dodecapeptides which contain oxidised glycine, valine, isoleucine, and
   tryptophan, and Omphalotin B was successfully isolated from the
   basidiomycete omphalotus olearius in the early phase ([92]Büchel et
   al., 1998). All peptides in differential metabolites exhibited
   down-regulation in OYM, such as Lys-Leu, Pro-Ile-Phe, Arg-Leu,
   omphalotin B; additionally, no significant advantage was observed in
   nutrient substances in OYM. According to the above results, we can not
   confirm whether organic yak milk has more nutritional value than
   non-organic yak milk. Differences in metabolites may be produced by
   different genetics and different regional environments, and the above
   results show that the metabolite compositions of OYM and NOYM are more
   similar compared to “OJM vs. NOJM”, which can be explained by the fact
   that both organic and non-organic feeding pastures for Yaks are located
   in highlands with harsh environments.

Table 3.

   The information of all potential differential metabolites from the
   pairwise comparison “OYM vs. NOYM”.
   ID Metabolite Formula M/Z Adducts Regulate VIP FC p-value
   1 Trimethylamine N-oxide C3H9NO 76.07601 M + H down 1.69 0.86 0.01
   2 Dihydroxyacetone C3H6O3 244.0785 2 M + ACN + Na down 1.10 0.96 0.01
   3 Deoxycytidine C9H13N3O4 228.0971 M + H down 1.18 0.93 0.05
   4 2′-Deoxyadenosine 5′-phosphate C10H14N5O6P 332.0741 M + H up 3.95
   5.29 0.01
   5 Capsicoside C1 C44H72O18 445.2426 M + 2H down 3.40 0.26 0.05
   6 Alpha-Methyltryptamine C11H14N2 175.1225 M + H up 1.79 1.15 0.00
   7 Hydroxykynurenine C10H12N2O4 207.0758 M + H-H2O up 3.50 2.71 0.03
   8 Tiglylglycine C7H11NO3 158.0807 M + H up 1.10 1.05 0.01
   9 PS (22:2 (13Z,16Z)/TXB2) C48H84NO14P 487.7691 M + 2Na down 3.94 0.00
   0.00
   10 PG (i-12:0/a-13:0) C31H61O10P 324.1963 M + H + Na down 1.02 0.95
   0.05
   11 Leu-Pro C11H20N2O3 211.1435 M + H-H2O down 1.17 0.94 0.03
   12 CDP-DG (a-15:0/20:5 (6E,8Z,11Z,14Z,17Z)–OH (5)) C47H77N3O16P2
   1015.495 2 M + 3H2O + 2H down 1.31 0.94 0.02
   13 l-Pyridosine C12H18N2O4 255.1331 M + H, M + Na, M + H-H2O,
   M + H-2H2O up 1.83 1.12 0.00
   14 Cyclic adenylic acid C10H12N5O6P 330.0585 M + H, M + Na up 1.06 1.04
   0.00
   15 Cer (d16:1/20:5 (7Z,9Z,11E,13E,17Z)–3OH (5,6,15)) C36H61NO6 586.4505
   M + H-H2O down 1.57 0.91 0.04
   16 Cer (d18:2(4E,14Z)/20:5 (7Z,9Z,11E,13E,17Z)–3OH (5,6,15)) C38H63NO6
   630.4767 M + H down 2.13 0.83 0.02
   17 Hericenone D C37H58O6 616.4607 M + NH4 down 1.22 0.95 0.03
   18 4-Oxo-1-(3-pyridyl)-1-butanone C9H9NO2 164.0701 M + H, M + NH4 down
   1.33 0.93 0.00
   19 PE (LTE4/P-18:1 (11Z)) C46H81N2O10PS 898.5479 2 M + 3H2O + 2H down
   1.69 0.87 0.02
   20 PS (18:1 (12Z)–2OH (9,10)/14:0) C38H72NO12P 405.7277 M + 2Na down
   1.45 0.89 0.04
   21 Pro-Ile-Phe C20H29N3O4 376.2217 M + H down 3.44 0.24 0.04
   22 [6]-Gingerdiol 4′-O-beta-d-glucopyranoside C23H38O9 491.2849
   M + CH3OH + H down 1.39 0.92 0.04
   23 Methyl 4-amino-5-ethyl-3-thiophenecarboxylate C8H11NO2S 186.0577
   M + H down 1.25 0.92 0.04
   24 PS (14:0/20:4 (8Z,11Z,14Z,17Z)–2OH (5S,6R)) C40H70NO12P 405.7277
   M + H + Na down 1.83 0.84 0.03
   25 Phe-Pro C14H18N2O3 263.1381 M + H down 1.06 0.95 0.02
   26 Salsolinol C10H13NO2 162.0923 M + H-H2O down 1.43 0.90 0.01
   27 Indole-3-Carboxylic Acid C9H7NO2 162.0545 M + H down 1.25 0.95 0.00
   28 CHEBI: 69,439 C10H16N2O2 197.1279 M + H down 1.21 0.95 0.05
   29 Omphalotin B C74H123N13O18 752.9563 M + H + Na, M + 2Na down 3.78
   0.00 0.00
   30 DOPA sulfate C9H11NO7S 242.011 M + H-2H2O down 1.06 0.95 0.02
   31 Pyridine N-oxide glucuronide C11H14NO7+ 314.1084 M + ACN + H down
   1.05 0.96 0.04
   32 Arg-Leu C12H25N5O3 288.2019 M + H down 1.83 0.81 0.03
   33 Lys-Leu C12H25N3O3 260.196 M + H down 3.56 0.22 0.03
   34 Draflazine C30H33Cl2F2N5O2 586.1934 M + H-H2O down 1.15 0.96 0.00
   [93]Open in a new tab

   In addition, a pathway enrichment analysis was performed on these
   differential metabolites to explain the relationship between variation
   in the chemical composition and dairy farming management systems. The
   enrichment background in KEGG annotation was all metabolites of Yak. We
   found that the differential metabolites between OYM and NOYM were
   involved in 80 metabolic pathways, of which 69 pathways were
   significantly enriched (p < 0.05) and the top 20 are shown in [94]Fig.
   3d. Most pathways involved the endocrine systems. It was revealed that
   four pathways with maximum enrichment degree, namely hedgehog signaling
   pathway, vasopressin - regulated water reabsorption, longevity
   regulating pathway-multiple species, and circadian rhythm. These
   results suggested that the organic feeding management system was found
   to possess a high impact on these metabolic pathways of Yak. Previous
   research has shown that the metabolic pathways of nucleic acid
   derivatives (purines and pyrimidines) of cow are significantly affected
   by feed composition ([95]Rocchetti et al., 2022). In this study, we
   found that purine and pyrimidine metabolism in Yak are also affected by
   the organic feeding system. Furthermore, we found that some metabolic
   pathways affected by organic feeding strategies are shared in Yak and
   Jersey, such as serotonergic synapse, pyrimidine metabolism, chemical
   carcinogenesis - DNA adducts, ABC transporters, purine metabolism,
   metabolism of xenobiotics by cytochrome P450, and glycerophospholipid
   metabolism. There is currently a lack of focus on the metabolomics of
   Yak milk, especially organic yak milk, and the results of this study
   could provide new insights into the metabolite composition of Yak milk.

3.4. Verification of markers using real milk samples

   The utility of potential markers was further determined using authentic
   OJM, NOJM, OYM, and NOYM samples. In this study, the content thresholds
   of three candidate markers were quantified. The peptide of Thr-Ala-Val
   had contents of 223.78–432.38 ng/g and 993.06–2137.86 ng/g in NOJM and
   OJM samples, respectively. Furthermore, D-biotin contents ranged from
   232.16 to 676.08 ng/g in NOJM samples and from 46.17 to 96.29 ng/g in
   OJM samples. It is very apparent that peptide of Thr-Ala-Val was
   determined to be a marker of OJM, while D-biotin was determined to be a
   marker of NOJM ([96]Fig. 4a, 4b). D-Biotin, also known as coenzyme R
   and vitamin H or B7, has been recognized as an essential nutrient; it
   participates as a cofactor in gluconeogenesis, fatty acid synthesis and
   branched chain amino acid catabolism ([97]Gravel & Narang, 2005).
   Previous research demonstrated that biotin was abundant in milk and our
   finding further indicated biotin was more abundant in NOJM than OJM. To
   further validate the detection sensitivity of markers, the adulterated
   simulants were prepared by mixing 10, 20, 30, and 60% NOJM in OJM. The
   content of D-biotin in the adulterated simulants did not fall in the
   range of that in OJM when the percentage of NOJM was up to 20%. Thus,
   as little as 20% adulteration of NOJM could be identified using
   D-biotin as the judgment index. If the percentage of NOJM in
   adulterated simulants exceeded 60%, the content of Thr-Ala-Val in
   adulterated simulants fell outside the scope of that in OJM.

Fig. 4.

   [98]Fig. 4
   [99]Open in a new tab

   Quantitative results for markers in authentic NOJM, OJM, NOYM, and OYM
   samples. a. Peptide of Thr-Ala-Val, b. D-biotin, and c. trimethylamine
   N-oxide.

   NOYM and OYM had a trimethylamine N-oxide of 175.07–223.21 and
   110.41–135.17 ng/g, respectively ([100]Fig. 4c). In dairy cows, dietary
   choline, betaine, and levocarnitine are all degraded by rumen
   microorganisms to trimethylamine, which is then endogenously
   biosynthesized by trimethylamine N-oxide ([101]Myers et al., 2021).
   Previous studies have linked trimethylamine n-oxide to the progression
   of cardiovascular disease, chronic kidney disease, fatty liver, and
   insulin resistance ([102]Janeiro, Ramírez, Milagro, Martínez, & Solas,
   2018). Our results showed that trimethylamine N-oxide was more abundant
   in NOYM samples. To further validate the detection sensitivity of
   markers, the adulterated simulants were prepared by mixing 10, 20, 30,
   and 60% NOYM in OYM. The content of trimethylamine N-oxide in
   adulterated simulants did not fall in the range of that in OYM when the
   percentage of NOYM was up to 30%. Thus, as little as 30% adulteration
   of NOYM could be identified using trimethylamine N-oxide as the
   judgment index.

   In brief, the D-biotin can effectively identify OJM samples adulterated
   with 20% NOJM, whereas peptide of Thr-Ala-Val was determined to be
   markers of OJM; and the trimethylamine N-oxide can effectively identify
   OYM samples adulterated with 30% NOYM. However, generally speaking, the
   adulteration rate is generally > 30%, even substituting non-organic
   milk as organic milk to obtain illegal profits in the market
   circulation because adulteration at lower levels of < 30% is associated
   with a smaller economic pay-off and is less often expected ([103]Kang
   et al., 2022b; [104]Wang et al., 2022). Hence, the use of these three
   markers were effective to determine the authenticity of Jersey and Yak
   organic milk.

4. Conclusion

   This study used untargeted and targeted metabolomics analysis to
   identify metabolite biomarkers for organic milk both Yak and Jersey.
   According to the differential analysis between metabolic profiles of
   organic and non-organic milk, 85 and 34 differential metabolites were
   identified as candidate molecular biomarkers in OYM and OJM,
   respectively. Finally, based on targeted metabolomics, peptide of
   Thr-Ala-Val and D-biotin were further confirmed as metabolite
   biomarkers for distinguishing OJM from NOJM, and trimethylamine N-oxide
   was confirmed as metabolite biomarkers for distinguishing OYM from
   NOYM. In the validation experiments, it was found that D-biotin can be
   used to detect adulteration of OJM with NOJM at levels as low as 20%
   and trimethylamine N-oxide has the ability to identify adulteration of
   OYM with NOYM at levels as low as 30%. In brief, this study
   demonstrated that the metabolomics analysis method was effective in
   distinguishing organic and non-organic milk. In summary, our findings
   provide new molecular markers to assess the authenticity of organic
   milk for both Jersey and Yak, and will facilitate the research on new
   detection methods.

CRediT authorship contribution statement

   Min Kang: Investigation, Data curation, Methodology, Writing – original
   draft. Hongxia Wang: Investigation, Methodology. Chuxin Chen:
   Investigation, Methodology. Ran Suo: Conceptualization, Methodology,
   Data curation, Supervision, Writing – review & editing. Jianfeng Sun:
   Conceptualization, Methodology, Data curation, Supervision. Quanhong
   Yue: Methodology. Yaqiong Liu: Validation.

Declaration of Competing Interest

   The authors declare that they have no known competing financial
   interests or personal relationships that could have appeared to
   influence the work reported in this paper.

Acknowledgements