Comparison and correlation analysis of microbiota composition and short-chain fatty acid profile in Milk across 15 mammalian species

Abstract

Milk contains a rich and diverse microbiota and short-chain fatty acids (SCFAs), both of which play important roles in offspring development. This study conducted a systematic cross-species comparative analysis of milk from 15 mammalian species using 16S rRNA sequencing and GC–MS techniques. The results revealed a conserved core microbiota at the phylum level, primarily composed of Firmicutes, Proteobacteria, and Actinobacteria, while at the genus level, Pseudomonas, Lactobacillus, and Propionibacterium were predominant. Host physiological traits and ecological environment were identified as key drivers of milk microbial composition. The specific enrichment of Rhizobium in human milk and Gemella in donkey milk may be related to maternal glucose metabolism, Stenotrophomonas and Jeotgalicoccus detected in multiple mammalian milk samples were closely associated with mammary gland health, and the dominant presence of Thermus in yak, camel, and buffalo milk reflects adaptation to plateau or warm climates. Significant differences were observed in the SCFA profiles among milk from different species. Human milk exhibited high levels of butyric acid (12.94 ± 7.63 μg/mL), valeric acid (0.12 ± 0.07 μg/mL), and caproic acid (11.08 ± 5.82 μg/mL). The SCFA profiles of Holstein cow milk and Mongolian horse milk were highly similar to those of human milk, particularly in butyric acid content (17.99 ± 5.15 μg/mL and 15.21 ± 4.89 μg/mL), suggesting their potential as alternative sources to human milk. A close interaction exists between microbes and SCFAs. Phyla such as Firmicutes and Verrucomicrobia can produce SCFAs through lactose fermentation. Genera including Actinomyces, Akkermansia, and Bifidobacterium showed positive correlations with various SCFAs, contributing to nutrient absorption and probiotic functions. In contrast, potentially pathogenic bacteria such as Actinobacillus, Corynebacterium, and Staphylococcus were negatively correlated with SCFAs, indicating that SCFAs may help maintain microbial balance by inhibiting pathogens. These findings demonstrate a bidirectional regulatory relationship between milk microbiota and SCFAs, which collectively contribute to the stability of the mammary gland microenvironment and newborn intestinal health.

The results not only provide new insights into the conservation and species-specificity of the milk microecology but also offer a important theoretical basis for developing human milk substitutes, functional dairy products, and intestinal health intervention strategies for infants and young children.

Highlights

• •This study systematically compared the microbiome and fatty acids in milk from various species.
• •This somewhat supports the possibility of a gut-mammary pathway.
• •It showed a link between milk microbiota and volatile fatty acids.

1. Introduction

Mammalian milk, serving as the core carrier of early nutrition and immunity for offspring, contains not only nutrients such as proteins and lipids but also harbors a dynamic microbial community and functional components, including short-chain fatty acids (SCFAs) (Lyons et al., 2020). SCFAs in breast milk primarily exist bound to triglycerides and are synthesized from the fermentation of oligosaccharides by anaerobic intestinal microbiota. This process may be accompanied by the transfer of intestinal microbes to mammary tissue via the immune cell-mediated gut-mammary axis (Xi et al., 2024). Milk microorganisms, acting as a source of the infant gut microbiota (Kalbermatter et al., 2021), contribute to the establishment of a microbial ecosystem during infant growth and development. These microbes, together with SCFAs, jointly regulate host physiological metabolism (Lyons et al., 2020) and enhance host immune capabilities, such as anti-infection responses (Xi et al., 2024).

Comparative studies have shown that although infant formula based on cow and caprine milk is supplemented with prebiotics to simulate breast milk function, such formula leads to delayed infant gut colonization due to reduced microbial diversity and imbalanced SCFAs, and significantly increases the risk of allergies and metabolic diseases (Ho et al., 2018). In contrast, breast milk is dominated by Bifidobacterium spp. and other beneficial bacteria, which synergistically inhibit the growth of pathogenic microorganisms in concert with SCFAs (Fusco et al., 2023). This highlights the importance of the milk microbiota-metabolite network. Recent studies have found that breast milk butyrate and valerate are significantly associated with Clostridium abundance (Xi et al., 2024). However, mammals have developed diverse lactation strategies and habitat adaptability during evolution, which may lead to differentiation in milk microbial structure and SCFA profiles among species. Currently, there is a lack of cross-species comparative data to determine whether the microbiota in the milk of economic animals (e.g., yak, camels, horses), and particularly its interaction with SCFAs, exhibits conservation or specificity among different species.

The selection of species in this study was based on multidimensional considerations to ensure ecological representation. Major dairy-producing animals, including Holstein cow, Saanen goat, and East Friesian sheep, were incorporated, along with regionally critical economic animals such as Qinghai yak, Alxa camel, and Guangxi buffalo, to investigate the influence of diverse habitats on milk composition. The selected species cover both ruminants and non-ruminants. Using 16S rRNA gene sequencing and GC–MS technology, we aim to establish the first multi-species milk microbiota–metabolite association dataset. This effort not only fills a critical knowledge gap in the milk microbiome research of economically significant animals but also provides a foundational reference for the development of functional dairy products and the optimization of SCFA profiles in infant formula.

2. Materials and methods

2.1. Sample collection

Milk samples were collected from fifteen mammalian species, Chinese human milk (CHP), Holstein cow milk (HST), Jersey cattle milk (JSC), Guangxi buffalo milk (GXB), Qinghai Yak milk (QHY), Dezhou donkey milk (DZD), Mongolian horse milk (MGH), Alxa camel milk (ALC), Saanen goat milk (SNG), Nubian goat milk (NBY), Toggenburg goat milk (TGB), Liaoning Cashmere goat milk (LCG), Small Tail Han sheep milk (STH), East Friesian sheep milk (EFS), and Yangxiang pig milk (YXP) (Table 1), with six biological replicates per species. All samples were collected from healthy individuals during autumn. For all species, milk samples were obtained during the second week of lactation following the birth of the second offspring. At the end of each feeding or milking session, milk samples were collected and immediately placed into individual sterile 10 mL plastic containers. Samples were then stored at −20 °C for short-term storage (≤ 2 weeks). Subsequently, samples were transported on dry ice and stored at −80 °C for long-term storage (≤ 4 months) without thawing. All sample collection and subsequent analyses were performed within the same experimental timeframe.

Table 1.

Table of Source Information and Abbreviations for 15 Milk Samples.

Samples	Sample Collection Site	Samples Abbreviations
Chinese human milk	Chinese mothers from Shenyang Maternal and Child Care Center	CHP
Holstein cow milk	Huishan Dairy Group	HST
Guangxi Buffalo milk	Guangxi Buffalo Research Institute and Nanjing Agricultural University	GXB
Qinghai Yak milk	Qinghai province and Lanzhou institute of husbandry and pharmaceutical science of cass	QHY
DeZhou Donkey milk	East Ajiao Donkey Farm in Fumeng County	DZD
Mongolian Horse milk	Inner Mongolia XilinGol League Science and Technology Association	MGH
Alxa camel milk	Alxa Alpaca Farm in Inner Mongolia	ALC
Saanen milk goat milk	Liaoyang dairy goat farm	SNG
Nubian milk goat milk	Liaoyang dairy goat farm	NBY
Toggenburg milk goat milk	Liaoyang dairy goat farm	TGB
Liaoning Cashmere goat milk	Liaoyang National Core Sheep Breeding Farm	LCG
Small Tail Han sheep milk	Inner Mongolia Horinger County sheep farm	STH
Yangxiang Pig milk	Yangxiang Pig Group	YXP
East Friesian sheep milk	Liaoyang dairy goat farm	EFS
Jersey cattle milk	Huishan Dairy Group	JSC

2.2. 16S rRNA Gene Variable Region Amplicon Sequencing

2.2.1. Microbial DNA extraction

Microbial DNA extraction was carried out using the Mo Bio/QIAGEN DNeasy PowerSoil Kit. The extracted DNA underwent thorough assessment. DNA concentration was determined employing a fluorescence spectrophotometer (Quantifluor-ST fluorometer, Promega, E6090) at both 260 nm and 280 nm to precisely measure DNA concentration. The quality of the DNA was further confirmed by subjecting it to 1 % agarose gel electrophoresis. DNA solutions were meticulously adjusted to achieve the desired concentration, with working DNA solutions being stored at 4 °C and storage solutions stored at −20 °C for long-term preservation.

2.2.2. PCR amplification

First, amplification of the 16S rRNA gene variable region was carried out as follows: (1) A preliminary PCR assay targeting the specific V region of the sample DNA was conducted. (2) Large-scale PCR amplification was performed using Pyrobest DNA Polymerase (TaKaRa, DR500A). Gel purification was conducted by excising and recovering the target bands to obtain purified samples (AxyPrep DNA Gel Extraction Kit, Axygen, AP-GX-500). Quantification of each sample was then performed using a BioTek microplate reader (BioTek FLx800 microplate reader, Quant-iT PicoGreen dsDNA Assay Kit, Invitrogen, P7589). Finally, the standard Illumina TruSeq DNA library preparation procedure (Illumina TruSeq DNA Sample Preparation Guide) was employed to construct the required sequencing libraries.

2.2.3. Library construction and sequencing workflow

In this section, we describe the general workflow for library construction and sequencing. DNA fragments were repaired, adapted, and enriched to generate sequencing libraries, which were then validated, normalized, and pooled prior to sequencing. Detailed protocols for each step are provided in the Supplementary Information section.

2.3. Extraction and detection of short-chain fatty acids

2.3.1. Extraction of short-chain fatty acids

Acetic acid, propionic acid, butyric acid, isobutyric acid, valeric acid, and isovaleric acid reference standards (purity ≥99.0 %, Supplementary Table 1.) were accurately weighed (100 mg each) and dissolved in ultrapure water to prepare stock solutions at a concentration of 100 mg/mL. A series of working standard solutions at varying concentrations were subsequently prepared by serial dilution for the construction of calibration curves. All stock solutions were stored at −20 °C, and working standard solutions were prepared fresh prior to use.

Take 100 μL of the sample should be taken and placed in a 2 mL centrifuge tube. Add 50 μL of 15 % phosphoric acid, followed by 10 μL of a 75 μg/mL internal standard solution (isovaleric acid) and 140 μL of ether. Mix the contents thoroughly for 1 min and then centrifuge at 12,000 rpm at 4 °C for 10 min. The sample is now ready for analysis.

2.3.2. GC–MS mass spectrometry

The gas chromatography system used was ThermoTrace 1300 (ThermoFisher Scientific, USA) with an Agilent HP-INNOWAX capillary column (30 m × 0.25 mm ID × 0.25 μm). Split injection was performed with an on-column injection volume of 1 μL and a split ratio of 10:1. The injector temperature was set at 250 °C, ion source temperature at 300 °C, and transfer line temperature at 250 °C.The temperature program for the GC oven was as follows: start at 90 °C, then ramp up at 10 °C/min to 120 °C, followed by a ramp of 5 °C/min to 150 °C, and finally a ramp of 25 °C/min to 250 °C, maintaining at 250 °C for 2 min. Helium was used as the carrier gas at a flow rate of 1.0 mL/min. For mass spectrometry, a Thermo ISQ 7000 mass spectrometer (ThermoFisher Scientific, USA) equipped with an electron impact (EI) ionization source was used. The instrument operated in the selected ion monitoring (SIM) mode with an electron energy of 70 eV.

2.4. Data processing and bioinformatics analysis

The raw data from high-throughput sequencing were initially screened based on sequence quality. After quality filtering, the raw sequences were demultiplexed and divided into libraries and samples using index and barcode information, with barcode sequences removed. QIIME2 was used for sample annotation. The obtained samples were subjected to detailed composition analysis, alpha diversity analysis, beta diversity analysis, multivariate statistical analysis, and KEGG pathway analysis at different taxonomic levels using R software and QIIME2. For the SCFA data, quantitative analysis, multivariate statistical analysis, differential SCFA analysis, and KEGG pathway analysis were conducted using R software.

3. Results

3.1. Milk microbial community structure

3.1.1. Diversity of the Milk microbiota

Analysis of 16S rRNA data from 15 mammalian milk sources identified 8 core amplicon sequence variants (ASVs) (Fig. 1A). At the phylum level, Proteobacteria, Firmicutes, and Actinobacteria predominated. Goat (SNG/NBY/TGB/LCG) and sheep milk (EFS/STH) exhibited similar microbial profiles, with Proteobacteria as the dominant phylum. At the genus level, 1303 taxa were detected, revealing distinct microbial communities across species. Core genera included Pseudomonas, Lactobacillus, and Propionibacterium. Streptococcus was the dominant genus in CHP, exhibiting the highest relative abundance. Pseudomonas was enriched in QHY, while Stenotrophomonas dominated HST (Fig. 1C).

Fig. 1.

(A) Venn flower diagram of ASVs/OTUs. The overlapping region in the center represents the ASVs/OTUs shared by all sample groups. The number in each region indicates the count of ASVs/OTUs contained in the corresponding sample. (B) Grouped box plot of alpha diversity indices. Each panel corresponds to one alpha diversity index. The number below each diversity index label indicates the P-value from the Kruskal–Wallis test (0.01 < P < 0.05:*, 0.001 < P < 0.01: **, P < 0.001:***). Together, these panels form a comprehensive evaluation system for microbial community alpha diversity. (C) Bar plot of ASV/OTU taxonomy. The top 10 taxa in terms of average abundance and count are shown (left: phylum level, right: genus level). (D) PCoA two-dimensional sample score plot. Differences between sample groups were analyzed using PCoA. Samples closer together in the ordination space indicate more similar community compositions. (E) Heatmap of microbial abundance clustered at the genus level. In the species abundance heatmap, red indicates higher abundance and blue indicates lower abundance of a genus in a sample. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Chao1 and Observed Species indices indicated that CHP had the lowest microbial richness (305.20 ± 50.17), whereas JSC (1791.12 ± 318.83) and DZD (1274.94 ± 395.95) showed higher richness. Integrated Shannon, Simpson, and Pielou indices revealed that LCG possessed the highest microbial diversity and evenness. Sequencing data completeness was robust (Good's coverage >0.9), providing a reliable foundation for downstream analyses (Fig. 1B).

β-diversity was evaluated via Principal Coordinates Analysis (PCoA), utilizing the weighted UniFrac distance. Microbial communities exhibited high structural similarity across species, with minimal intergroup variation. EFS and QHY formed tight intragroup clusters but showed significant interspecies divergence from other milk types (Fig. 1D).

3.1.2. Species difference analysis

Linear Discriminant Analysis Effect Size (LEfSe) was used to detect microbial taxa with significant enrichment among groups, with the screening criteria set as LDA ≥4 and P < 0.01. Streptococcus was enriched in CHP, while Staphylococcus exhibited higher abundance in TGB. Bifidobacterium and Lactobacillus were significantly abundant in JSC and GXB, respectively, suggesting their potential as biomarkers (Fig. 2A,B) (Supplementary Table 2.).

Fig. 2.

(A) Taxonomic branch diagram, showing microbial taxa from phylum to genus level. Green nodes represent taxa with no significant differences between groups, while colored nodes indicate significant intergroup differences (P < 0.05). (B) Bar chart of LDA effect sizes. The y-axis shows taxa exhibiting differences, with an LDA score ≥ 4 and P < 0.05. The length of each bar represents the effect size of each taxon, with longer bars indicating greater significance in differential abundance between groups. This method aids in identifying biomarker taxa with significant differences among groups. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Hierarchical clustering analysis at the genus level revealed distinct intersample microbial community patterns and abundance distribution patterns. CHP samples exhibited significantly elevated relative abundances of Streptococcus, Rhizobium (showing specific enrichment), Rothia, Bacillus, Nocardioides, Propionibacterium, Kaistobacter, and Planomicrobium compared to other milk sources. JSC demonstrated unique high-abundance features dominated by Psychrobacter and Ornithinicoccus. Species-specific enrichment patterns were observed across samples: TGB milk was characterized by Lawsonia dominance, while STH milk showed preferential enrichment of Agrobacterium and Brachybacterium. Prevotella served as the signature genus in LCG samples, whereas DZD milk was distinguished by Gemella, Sphingomonas, and Sphingobium. ALC samples featured Moraxella and Yaniella, EFS was typified by Chryseobacterium, and YXP exhibited Cupriavidus and Curvibacter as dominant taxa. Deinococcus was enriched in SNG, Stenotrophomonas dominated HST, and QHY samples showed Leuconostoc and Thermus as prevalent genera. Cross-species analysis further identified Acinetobacter enrichment in JSC, STH, NBY, MGH, EFS, and YXP samples. Notably, Corynebacterium and Arthrobacter demonstrated consistent enrichment patterns in ruminant-derived milks, indicating their role as conserved dominant taxa across these species (Fig. 1E).

3.1.3. Microbial co-occurrence and functional analysis

To further elucidate the overall microbial community structure in milk, we performed microbial co-occurrence network analysis (|r| > 0.5, P < 0.01). At the genus level, widespread correlations indicated predominantly synergistic relationships. 61 genera primarily affiliated with Proteobacteria, Bacteroidetes, Firmicutes, Actinobacteria, Tenericutes, and Thermi were enriched. Pseudomonas, Acinetobacter, and Lactobacillus exhibited prominent connectivity within the network, thereby identifying these taxa as keystone groups governing milk microbial community structures across species (Fig. 3A).

Fig. 3.

(A) Microbial co-occurrence network diagram. Nodes are colored according to their respective phyla, and their sizes are proportional to taxonomic abundance. Edges represent correlations between microorganisms. (B) Metabolic pathway statistics based on KEGG PATHWAY classification, showing the distribution of Count and Relative Abundance.

Prediction of microbiome metabolic pathways was performed using the KEGG database, with the number and relative abundance of secondary functional pathways summarized. The microbiota was primarily enriched in Metabolism, which contained the largest number of sub-pathways (125) and exhibited higher relative abundance compared to other functional categories, indicating its role as the core functional module of the microbial community. Among these, Carbohydrate metabolism and Amino acid metabolism showed the highest relative abundance, followed by Metabolism of cofactors and vitamins, and Metabolism of terpenoids and polyketides, suggesting that microbial communities may produce key metabolites through pathways such as carbohydrate decomposition and amino acid metabolism. In contrast, the Organismal Systems pathway demonstrated low relative abundance and a limited number of sub-pathways (11), implying a relatively minor direct regulatory influence on the samples. Additionally, pathways related to Cellular Processes, Environmental Information Processing, Genetic Information Processing, and Human Diseases were also identified. Together, these pathways form an integrated functional network of the microbial community (Fig. 3B).

3.2. Composition of SCFA in Milk

3.2.1. Quantitative analysis of SCFAs

Quantitative analysis of 7 SCFAs revealed significant variations among species (Table 2). Milk samples from GXB (21.31 ± 5.87 μg/mL) and DZD (18.13 ± 12 μg/mL) exhibited significantly higher acetic acid concentrations compared to other species (P < 0.01). Propionic acid concentrations were significantly elevated in GXB (2.18 ± 1.55 μg/mL) relative to all other species (P < 0.01). CHP samples contained higher concentrations of butyric acid (12.94 ± 7.63 μg/mL), valeric acid (0.12 ± 0.07 μg/mL), and caproic acid (11.08 ± 5.82 μg/mL). HST cow milk displayed significantly higher butyric acid (17.99 ± 5.15 μg/mL) and valeric acid (0.18 ± 0.08 μg/mL) concentrations than other species (P < 0.01). MGH samples demonstrated significantly elevated levels of isobutyric acid (0.34 ± 0.11 μg/mL) and isovaleric acid (1.00 ± 0.28 μg/mL) compared to other species (P < 0.01). Overall SCFA concentrations were most similar between CHP and HST milk samples (Fig. 4A).

Table 2.

Quantitative Results of SCFAs.

SCFAs (μg/mL)	CHP	YXP	SNG	NBY	TGB	LCG	EFS	STH	HST	JSC	GXB	QHY	ALC	MGH	DZD
Acetic acid	2.35 ± 0.43B	4.07 ± 1.35B	2.19 ± 0.66B	2.49 ± 0.67B	2.75 ± 0.38B	4.44 ± 1.69B	4.41 ± 0.66B	4.11 ± 0.78B	2.95 ± 0.67B	3.28 ± 1.24B	21.31 ± 5.87 A	3.73 ± 1.08B	4.81 ± 0.51B	4.15 ± 0.69B	18.13 ± 12 A
Propionic acid	0.06 ± 0.02B	0.03 ± 0.02B	0.05 ± 0.06B	0.09 ± 0.04B	0.14 ± 0.01B	0.08 ± 0.03B	0.1 ± 0.07B	0.18 ± 0.07B	0.22 ± 0.07B	0.2 ± 0.16B	2.18 ± 1.55 A	0.13 ± 0.05B	0.14 ± 0.04B	0.21 ± 0.06B	0.18 ± 0.11B
Isobutyric acid	0.03 ± 0.01C	0.01 ± 0C	0.01 ± 0.01C	0.01 ± 0C	0.01 ± 0C	0.02 ± 0.01C	0.07 ± 0.01BC	0.03 ± 0.01C	0.04 ± 0.01C	0.04 ± 0.01C	0.1 ± 0.05B	0.03 ± 0.01C	0.02 ± 0C	0.34 ± 0.11 A	0.06 ± 0.03BC
Butyric acid	12.94 ± 7.63AB	0.01 ± 0.01F	1.12 ± 1.1EF	1.06 ± 0.85EF	0.89 ± 0.63EF	3.36 ± 2.19DEF	1.63 ± 0.08EF	2.18 ± 0.49EF	17.99 ± 5.15 A	6.83 ± 2.69CDE	8.9 ± 3.43BCD	4.02 ± 1.46DEF	0.09 ± 0.02F	11.23 ± 6.72BC	2.57 ± 3.4EF
Isovaleric acid	0.03 ± 0.01C	0.05 ± 0.03C	0.01 ± 0C	0.01 ± 0C	0.01 ± 0C	0.01 ± 0C	0.02 ± 0C	0.01 ± 0C	0.02 ± 0.01C	0.02 ± 0.01C	0.02 ± 0.01C	0.02 ± 0.02C	0.02 ± 0C	1.00 ± 0.28 A	0.25 ± 0.31B
Valeric acid	0.12 ± 0.07B	0.01 ± 0E	0.01 ± 0.01E	0.01 ± 0.01E	0.01 ± 0.01E	0.02 ± 0.01DE	0.01 ± 0E	0.01 ± 0E	0.18 ± 0.08 A	0.07 ± 0.05BCD	0.08 ± 0.04BC	0.03 ± 0.02CDE	0.01 ± 0E	0.07 ± 0.04BCD	0.02 ± 0.01DE
Caproic acid	11.08 ± 5.82 A	0.03 ± 0.02C	0.76 ± 0.85C	0.94 ± 0.81C	0.65 ± 0.37C	2.14 ± 1.3C	0.97 ± 0.07C	0.95 ± 0.15C	7.16 ± 2.2B	2.46 ± 1.09C	2.48 ± 1.15C	1.52 ± 0.09C	0.03 ± 0.01C	8.99 ± 5.33AB	2.68 ± 3.95C

Note: Within each SCFA type, significant differences (P < 0.01) in concentration (mean ± SD) among milk samples are denoted by superscript letters A, B, C, D, E, F.

Fig. 4.

(A) Quantitative bar chart of SCFA concentrations. Within each SCFA type, significant differences (P < 0.01) in concentration (mean ± SD) among milk samples are indicated by superscript letters A, B, C, D, E, and F. (B) Principal component analysis (PCA) of SCFAs. The score plot visualizes the clustering and dispersion patterns of samples based on the composition and abundance of SCFAs.

3.2.2. Multivariate statistical analysis

Principal Component Analysis (PCA) revealed variations within and between groups, with the cumulative contribution rate of the principal components reaching 73.3 % (Fig. 4B). Relatively significant differences were observed among the ALC, GXB, and MGH groups. In contrast, the HST, JSC, QHY, LCG, and CHP groups exhibited minimal intergroup differences, high reproducibility, and similar molecular composition and concentration. Considerable divergence was detected between the CHP and bovine-related milk groups (HST, JSC, GXB, QHY) and the pig milk group (R²X = 0.806). The ALC and YXP groups demonstrated minimal separation along the PC2 axis (R²X = 0.8). Samples from goat and sheep milk groups clustered closely in the mid to negative region of PC1 and the negative region of PC2 (R²X = 0.685).

3.2.3. Screening of differential SCFAs

To identify statistically significant SCFAs, additional screening was performed using P < 0.05 and a variable importance in projection (VIP) score > 1. Metabolites meeting these criteria were selected for further analysis. Z-scores were calculated to evaluate SCFA relative content levels across species using the formula: z = (x - μ) / σ, where x represents the specific SCFA content in a sample, μ is the mean, and σ is the standard deviation. DZD milk exhibited a markedly high Z-score for acetic acid, consistent with its significantly elevated mean concentration (18.13 μg/mL). GXB displayed a high Z-score for propionic acid, supported by its notably high mean abundance (2.18 μg/mL). MGH showed a high Z-score for isobutyric acid, reflecting its uniquely high mean concentration (0.34 μg/mL).

To investigate associations among differentially expressed SCFAs, a correlation heatmap was generated. Among the 15 species studied, metabolic correlations for the five differential SCFAs were predominantly positive. Notably, butyric acid and valeric acid exhibited a strong positive correlation (r = 0.93, P < 0.05), suggesting potential synergistic biological functions. Furthermore, propionic acid showed positive correlations with all other differential SCFAs (P < 0.05) (Fig. 5B).

Fig. 5.

(A) Z-score plot of differential SCFA levels. Using CHP as the control, the relative abundance of each SCFA was normalized via Z-score. A position further to the right indicates higher content of the SCFA in that sample. (B) Correlation heatmap of differential SCFAs. Statistical testing was applied to assess associations among SCFAs in milk. Red indicates a positive correlation, with darker shades representing greater significance (P < 0.05). (C) Multi-group volcano plot of SCFA differences. Using CHP as the control, pairwise comparisons reflect both the magnitude (log_₂FC) and significance of differences in SCFAs. (Red P < 0.05, blue P ≥ 0.05). log_₂FC > 0 represents up-regulation in the treatment group compared to the control, while log_₂FC < 0 indicates down-regulation. (D) KEGG metabolite molecular network. Green nodes represent metabolic pathways, red nodes represent SCFAs. Larger node size corresponds to a greater number of associated metabolites or pathways. (E) Microbial–SCFA association network. The network was constructed based on Spearman's rank correlation coefficient (|r| > 0.2) calculated from abundance data across all milk samples. Red nodes represent bacterial phyla, and blue nodes represent SCFAs. Solid orange edges indicate positive correlations, while dashed purple edges represent negative correlations. Edge thickness corresponds to the strength of the correlation. (F) Redundancy analysis (RDA). The top 20 microbial genera with correlation coefficients significant at P < 0.05 are displayed. Acute angles between variable vectors indicate positive correlations (synergy), while obtuse angles denote negative correlations (antagonism). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Differential metabolite analysis, using CHP as the control, revealed significant interspecies variations in SCFA levels. Compared to CHP, DZD, HST, and MGH exhibited upregulation of four SCFAs. All other samples showed a predominance of downregulated SCFAs over upregulated ones. ALC and YXP demonstrated significantly greater downregulation relative to CHP (Fig. 5C).

3.2.4. SCFA pathway analysis

To elucidate the metabolic pathways associated with differential levels of short-chain fatty acids (SCFAs), KEGG enrichment analysis was performed. A total of 24 pathways were significantly enriched (P < 0.05). Five differential SCFAs—acetic acid, propionic acid, butyric acid, isovaleric acid, and isobutyric acid—were associated with the most enriched pathway, protein digestion and absorption. Acetic, propionic, and butyric acids were involved in the second most enriched pathway, carbohydrate digestion and absorption. Among these, acetic acid was linked to the highest number of metabolic pathways, participating in nine significantly enriched pathways: Protein digestion and absorption, Carbohydrate digestion and absorption, Propanoate metabolism, Degradation of aromatic compounds, Glycosaminoglycan biosynthesis - heparan sulfate / heparin, Cholinergic synapse, Taurine and hypotaurine metabolism, Glycolysis / Gluconeogenesis, Pyruvate metabolism (Fig. 5D).

3.3. Microbial-SCFA correlation analysis

The bacterial phyla common across multiple milk species included Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Cyanobacteria, Firmicutes, Proteobacteria, and Verrucomicrobia. Spearman correlation analysis revealed that microbial phyla were predominantly negatively correlated with SCFAs. Specifically, Acidobacteria and Cyanobacteria showed negative correlations with more than six SCFAs, while Actinobacteria, Proteobacteria, and Verrucomicrobia exhibited positive correlations with more than five SCFAs. Among these, the abundance of Verrucomicrobia was positively correlated with acetic acid and propionic acid (|r| > 0.3, P < 0.05). In contrast, a significant moderate negative correlation was observed between the abundance of Chloroflexi and acetic acid (|r| > 0.3, P < 0.01). Additionally, both Chloroflexi and Acidobacteria were negatively correlated with isobutyric acid (|r| > 0.2, P < 0.05) (Fig. 5E).

Furthermore, 32 common genera were identified. Staphylococcus, Turicibacter, Actinobacillus, Acinetobacter, Microbacterium, Chryseobacterium, Corynebacterium, Arthrobacter, Mycobacterium, Haemophilus, Kocuria, and Ruminococcus displayed varying degrees of negative correlation with valeric acid, butyric acid, and caproic acid. Conversely, Rothia, Phascolarctobacterium, Actinomyces, Enterococcus, Coprococcus, Lactobacillus, Propionibacterium, Bifidobacterium, Blautia, Roseburia, and Pseudomonas exhibited varying degrees of positive correlation with valeric acid, butyric acid, and caproic acid. Phascolarctobacterium, Clostridium, Coprococcus, Kocuria, Ruminococcus, and Akkermansia showed varying degrees of positive correlation with acetic acid, isobutyric acid, and propionic acid. The genera Actinomyces (Actinobacteria), Enterococcus (Firmicutes), and Lactobacillus (Firmicutes) were positively correlated with all SCFAs. Conversely, Staphylococcus (Firmicutes), Arthrobacter (Actinobacteria), and Shigella (Proteobacteria) were negatively correlated with all SCFAs. These taxa collectively participate in SCFA biosynthesis or co-metabolic regulation (Fig. 6).

Fig. 6.

Heatmap of Spearman correlation between SCFAs and the microbiome at the genus level. Orange indicates a positive correlation, purple indicates a negative correlation. Darker colors represent stronger correlations, while lighter colors indicate weaker correlations. Significance levels are denoted as follows: * 0.01 < P < 0.05, ** P < 0.01. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To explore the correlations and statistical significance between microbial communities and SCFAs, redundancy analysis (RDA) was performed at the genus level. The analysis focused on the top 20 genera contributing to the RDA (P < 0.05). RDA1 and RDA2 together explained 30.09 % of the total variation, indicating a discernible environmental influence on species distribution. Propionic acid (RDA1, |r| > 0.5, P < 0.05) was positively correlated with Lactobacillus, Streptococcus, and Actinomyces, but negatively correlated with Pseudomonas, Phascolarctobacterium, and Roseburia, suggesting that propionic acid may be a key driver shaping microbial distribution along the RDA1 axis. Acetic acid (|r| > 0.5, P < 0.05) represented the second strongest influence on RDA1. RDA2 captured secondary sources of variation. Isobutyric acid and isovaleric acid were positively correlated with genera along RDA2 (Phascolarctobacterium, Enterococcus, Oscillospira), potentially reflecting the capacity of these microbes to utilize branched-chain SCFAs and their adaptation to the corresponding growth environment. Isobutyric acid and isovaleric acid clustered together, while acetic acid, propionic acid, and butyric acid formed a separate cluster, indicating potential synergistic effects among these SCFAs in driving microbial community shifts (Fig. 5F).

4. Discussion

The structure of the microbial community in milk is influenced by nutritional components such as SCFAs (Xi et al., 2024). Unlike previous studies primarily focused on human milk, this research systematically analyzed the association between microbial communities and SCFA composition in milk from 15 mammalian species.

4.1. Characteristics of microbial community composition

Analysis of multi-species milk samples confirmed the richness and diversity of microbial communities. At the phylum level, Firmicutes, Proteobacteria, and Actinobacteria constituted the core phyla (Xi et al., 2024). At the genus level, Pseudomonas, Lactobacillus, Propionibacterium, among others, were ubiquitous across all species, collectively forming the core milk microbiota. The distribution of certain genera, however, may be influenced by host metabolism and external factors (Yatsunenko et al., 2012), exhibiting significant species-specific enrichment patterns. For instance, Lactobacillus, Streptococcus, Rhizobium, Rothia, Bacillus, Nocardioides, Propionibacterium, and Planomicrobium were significantly enriched in human milk. Notably, Rhizobium was specifically enriched in healthy human milk. Previous studies indicate higher abundance of Rhizobium in milk from mothers with gestational diabetes mellitus (GDM) (Li et al., 2025), suggesting its abundance may be linked to maternal glucose metabolism status (e.g., GDM). Similarly, Gemella was specifically enriched in donkey milk, its abundance regulated by maternal blood glucose levels (LeMay-Nedjelski et al., 2020). The high lactose content (∼6.5 %) in donkey milk (Xie et al., 2025) may provide a selective growth advantage for Gemella. These findings highlight the regulatory role of host metabolism on milk microbiota.

Changes in the abundance of certain specific genera are also closely linked to mammary health. For example, Stenotrophomonas was specifically enriched in cow milk, while Jeotgalicoccus was commonly found in goat, sheep, and camel milk (this study is the first to report its colonization in pig, horse, and donkey milk) (Biçer et al., 2021). In healthy mammary glands, these microbes can effectively inhibit the growth of pathogens like Staphylococcus aureus through metabolism. Conversely, microbial dysbiosis is often associated with the co-participation of pathogenic bacteria like S. aureus in the development of mastitis (He et al., 2023). Therefore, monitoring changes in the abundance of specific milk microbial biomarkers could contribute to establishing a risk surveillance system for mammary health.

The origins of milk microbiota are diverse. Several typical gut-associated genera, such as Faecalibacterium, were detected across multiple species. The presence of Psychrobacter and Ornithinicoccus in JSC milk and Lawsonia in TGB milk provides supporting evidence for the “gut-mammary axis” hypothesis (Pang et al., 2018). Some environmental microbes, such as Cupriavidus and Curvibacter, were enriched in pig milk and present at low abundance in human, cow, goat, sheep, buffalo, and camel milk, potentially linked to Pseudomonas metabolism (Le et al., 2023). Thermus, specifically enriched in yak milk, was also found in camel and buffalo milk (Zhao et al., 2020), its distribution possibly related to the host's high-altitude or thermophilic habitat. These findings suggest that microbial interactions within milk or environmental exposure are significant factors in the colonization of species-specific genera.

Furthermore, some species-specific genera exhibit unique dairy application value. Brachybacterium, significantly enriched in Small Tail Han sheep milk, possesses halophilic properties conferring a natural advantage for Gouda cheese production, generating distinctive flavor compounds while inhibiting spoilage bacteria (Salazar et al., 2018). Yaniella, a thermophilic spore-forming bacterium in Alxa camel milk, could serve as a hygiene indicator for pasteurized dairy products (Alexa et al., 2024). Similarly, Deinococcus in Saanen goat milk is typically only present in raw milk (Niyazbekova et al., 2020), making it a potential biomarker for distinguishing raw from pasteurized milk. These discoveries lay the foundation for utilizing microbial resources and enhancing quality control in dairy processing.

4.2. SCFA profile characteristics and potential functions

Detection of SCFAs uncovered marked differences across species. Human milk was rich in butyric, valeric (12.94 ± 7.63 μg/mL), and caproic acids (11.08 ± 5.82 μg/mL), with concentrations consistent with known studies (Barman et al., 2024; Stinson et al., 2020). Milk SCFAs serve as key drivers for infant gut microbiota colonization, providing the initial SCFA supply to the neonatal intestine (Li et al., 2025). They promote infant health and development through multiple mechanisms (e.g., as an energy source for intestinal epithelial cells, epigenetic regulation of immunity) (Paparo et al., 2021). Further studies indicate that the cumulative effects of butyrate and valerate may enhance SCFA inhibition of histone deacetylases (HDACs), potentially linking them to breast cancer development (Muradás et al., 2024; Stinson et al., 2020). Due to their limited quantity in human milk and primary role in offspring development, human milk SCFAs cannot be directly used as raw material for product development. However, high concentrations of butyric, valeric, and caproic acids were detected in Holstein cow milk and Mongolian horse milk. These may possess functional potential similar to SCFAs in human milk. Subsequent research could explore developing human milk substitutes based on cow or horse milk for applications in intestinal immune modulation and adjunctive cancer therapy. Furthermore, isobutyric acid (0.34 ± 0.11 μg/mL) and isovaleric acid (1.00 ± 0.28 μg/mL) levels in MGH were significantly higher than in other species. Their specific functions warrant further investigation and may offer new directions for the differentiated development of substitute milks. GXB (21.31 ± 5.87 μg/mL) and DZD (18.13 ± 12 μg/mL) were characterized by high acetic acid content, while ALC (4.81 ± 0.51 μg/mL) and YXP (4.07 ± 1.35 μg/mL) exhibited highly similar SCFA profiles dominated by acetic acid. Given that acetic acid has the potential to inhibit hepatic gluconeogenesis and lower blood glucose (Pyo et al., 2022), and inosine, highly expressed in both ALC and YXP, has also been shown to participate in blood glucose regulation (Wang, Sun, et al., 2023), high-acetic-acid milk types may be suitable for special populations such as diabetics (Li et al., 2020). However, whether these milks share more compositional commonalities and the underlying metabolic mechanisms require further in-depth research.

4.3. Interactions between Milk microbiota and SCFAs

Milk microbial communities (Actinobacteria, Bacteroidetes, Firmicutes, Proteobacteria, Verrucomicrobia, etc.) are highly similar to the infant gut microbiota and serve as a source for neonatal gut colonization (De Vadder et al., 2014; Solís et al., 2010). These microbes generate SCFAs through lactose fermentation, proteolysis, and metabolic synergy (Solís et al., 2010). Support for the acid-producing capacity of milk-derived microbes comes from Alothman et al. (Alothman et al., 2017), who observed a significant increase in SCFA concentration in cow milk inoculated with Pseudomonas fluorescens and Aureobacterium.

Based on this, our study assessed the relationship between milk microbiota and SCFAs. Firmicutes in milk showed a positive correlation with acetic acid, contrasting with the butyrate-producing dominance of this phylum in the gut. This difference may stem from the predominance of facultative anaerobes in the milk environment, suggesting that oxygen levels may influence SCFA composition by regulating microbial metabolic pathways (Xi et al., 2024).

Within Firmicutes, both Enterococcus and Lactobacillus correlated positively with SCFAs. Enterococcus is common in raw milk and fermented dairy products. Studies show E. faecium can modulate SCFA production through metabolic synergy with Akkermansia, Lactobacillus, and Bifidobacterium (Xu et al., 2022). Furthermore, Enterococcus strains isolated from human milk possess probiotic potential, capable of intervening in cholesterol metabolism and lowering host cholesterol levels (Bagci et al., 2019). Microbiota like Lactobacillus can break down indigestible nutrients in milk, providing galactose as a carbon source for Bifidobacterium, indirectly promoting enzyme synthesis and enhancing infant nutrient absorption efficiency (Fusco et al., 2023; Xi et al., 2024). Akkermansia within Verrucomicrobia, showed significant positive correlations with acetic (P < 0.01) and propionic acids (P < 0.05). This bacterium primarily relies on milk oligosaccharides as a specific carbon source and is a biomarker for gut health (Hagi & Belzer, 2021). Exogenous supplementation of Akkermansia muciniphila enhances intestinal epithelial development (Zhang et al., 2021).

Prevotella correlated positively with acetic acid generation. Acetic acid is a major precursor for milk fat synthesis. Prevotella in milk may promote milk fat synthesis by maintaining local acetic acid concentrations in the mammary gland. However, excess Prevotella might inhibit propionic acid production, affecting mammary cell uptake of other nutrients, which relates to energy metabolism during lactation (Wang, Chen, et al., 2023). Interestingly, Propionibacterium, traditionally considered a propionate producer, exhibited negative correlations with both acetic and propionic acid (P < 0.01) abundance in this study, the specific mechanisms of which require elucidation (Jan et al., 2002).

While Clostridium and Faecalibacterium are known to promote butyrate generation (Jia et al., 2018), this study further identified Phascolarctobacterium, Blautia (P < 0.05), Roseburia and Pseudomonas (P < 0.01) as showing significant positive correlations with butyric acid concentration in milk. Conversely, Ruminococcus correlated negatively with butyric acid but strongly positively with acetic acid (P < 0.01). We speculate that under substrate limitation or metabolic stress in milk, acid kinase activity might be inhibited in some Ruminococcus species, leading to the production of acetate, propionate, or amines instead of butyrate (Prentice et al., 2019).

We observed that Staphylococcus, Arthrobacter, and Shigella correlated negatively with all SCFAs. Corynebacterium and Aerococcus correlated negatively with SCFAs other than acetate, while Stenotrophomonas correlated negatively with SCFAs other than valeric acid. SCFAs exhibit antimicrobial activity against potential pathogens, and their abundance is associated with healthy mammary glands (Coates et al., 2023). Therefore, this prevalent negative correlation pattern underscores the role of SCFAs in maintaining milk microbiota balance.

Microbial-SCFA interactions are crucial for offspring health. As bacterially derived metabolites, SCFAs exert physiological effects by activating G protein-coupled receptors, and changes in their composition may influence breast cancer risk (Freitas et al., 2020). Upon ingestion via breastfeeding, SCFAs synergize with milk microbes to drive infant gut microbiota colonization: they lower intestinal pH, promote mucus production, and create favorable conditions for beneficial bacteria colonization. They also activate immune regulation (e.g., suppressing pro-inflammatory cytokine release) (O'Keefe, 2016), reducing the risk of neonatal microbiota dysbiosis. Studies indicate that breastfed infants have significantly higher abundances of Bifidobacterium and Lactobacillus, along with higher SCFA levels in their intestines, compared to formula-fed infants, which is closely linked to infant health outcomes, such as reduced risk of celiac disease (Moawad et al., 2023).

Our analysis of multi-species milk microbiota composition supports the notion that microbial activity directly influences SCFA concentrations in milk. In turn, SCFAs feedback to regulate microbial community structure (Andrews et al., 2020), establishing a symbiotic relationship. Limitations remain: the relatively small sample size may restrict generalizability, and the study lacked in-depth functional gene or metabolic pathway analysis. Future research should expand the sample size for validation and integrate metagenomic / transcriptomic approaches with in vitro experiments to elucidate the underlying interaction mechanisms.

5. Conclusion

In summary, comparative analysis of 15 animal milk samples revealed significant bidirectional interactions between milk-derived microorganisms and short-chain fatty acids (SCFAs) (|r| > 0.5, P < 0.05), forming an interaction network. These interactions collectively contribute to maintaining mammary gland homeostasis and facilitate infant gut colonization and immune regulation. Although a conserved core microbiota (including Pseudomonas, Lactobacillus, and Propionibacterium) exists across milk from different species, the microbial community structure and SCFA composition are still influenced by host genetic background and ecological factors, exhibiting both diversity and specificity. For instance, Rhizobium was enriched in CHP, Gemella in DZD, Brachybacterium in STH, and Yaniella in ALC. Human milk was rich in butyric, valeric, and caproic acids, whereas Guangxi buffalo milk and Dezhou donkey milk were characterized by high acetate levels. These findings not only provide supporting evidence for the “gut-mammary axis” hypothesis but also offer a theoretical foundation and microbial resources for milk origin tracing, monitoring host metabolic status, developing infant formula and functional dairy products, and enhancing milk quality inspection.

The interaction between microorganisms and SCFAs constitutes an intrinsic regulatory mechanism within the mammary microecology: this mechanism not only synergistically promotes the accumulation of SCFAs through microbial metabolism but also modulates the microbial community structure via SCFA-mediated feedback, thereby inhibiting potential pathogens and maintaining mammary health. This study elucidates the systemic characteristics and functional potential of the milk ecosystem at a macro level and provides candidate strains and target microbial groups for subsequent mechanistic research. However, the specific molecular mechanisms underlying the interactions between microorganisms and SCFAs require further validation through expanded sample sizes and integrated multi-omics technologies to facilitate the translation of related theories into applications in dairy processing and infant nutritional intervention.

The following are the supplementary data related to this article.

Supplementary Table 1

Information of short-chain fatty acid standard substances.

Supplementary Table 2

Top bacterial genera in 15 mammalian milks by LDA.

Ethics approval and consent to participate.

Ethical approval for the involvement of human subjects in this study was granted by Shenyang Agricultural University Research Ethics Committee, Reference number 202106015. The collection of human milk samples was conducted with the informed consent of the individuals and their family members. The sample collection was carried out with the assistance of a medical professional. The collection of animal samples was performed with the consent of the farm owner and under the guidance of professional technicians to minimize animal stress responses.

CRediT authorship contribution statement

Yuan Pan: Writing – original draft, Data curation, Conceptualization. Yanzhi Wu: Writing – original draft, Data curation. Yinggang Sun: Data curation, Conceptualization. Rui Chen: Data curation. Qian Li: Data curation. Xiaowei Wang: Data curation. Qiu Zhang: Data curation. Siyi Li: Data curation. Zeying Wang: Writing – review & editing, Conceptualization.

Funding

Our scientific research was financially aided by three projects: 1. National Key Research and Development Program of China (2023YFD1300200).

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.

Data availability

Data will be made available on request.