Discrepancies among healthy, subclinical mastitic, and clinical mastitic cows in fecal microbiome and metabolome and serum metabolome

Mastitis is generally considered a local inflammatory disease caused by the invasion of exogenous pathogens and resulting in the dysbiosis of microbiota and metabolites in milk. However, the entero-mammary pathway theory may establish a possible link between some endogenous gut bacteria and the occurrence and development of mastitis. In the current study, we attempted to investigate differences in the gut microbiota profile and metabolite composition in gut and serum from healthy cows and those with subclinical mastitis and clinical mastitis. Compared with those of healthy cows, the microbial community diversities in the feces of cows with subclinical mastitis (SM) and clinical mastitis (CM) were lower. Lower abundance of Bifidobacterium , Rom-boutsia , Lachnospiraceae_NK3A20_group , Coprococcus , Prevotellaceae_UCG-003 , Ruminococcus , and Alistipes , and higher abundance of the phylum Proteobacteria and the genera Escherichia-Shigella and Streptococcus were observed in CM cows. Klebsiella and Paeniclostridium were significantly enriched in the feces of SM cows. Several similarities were observed in feces and serum metabolites in mastitic cows. Higher levels of proinflammatory lipid products (20-trihydroxy-leukotriene-B4, 13,14-dihydro-15-keto-PGE 2 , and 9,10-dihydroxylin-oleic acids) and lower levels of metabolites involved in secondary bile acids (deoxycholic acid, 12-ketolithocho-lic acid), energy (citric acid and 3-hydroxyisovalerylcar-nitine), and purine metabolism (uric acid and inosine) were identified in both SM and CM cows. In addition, elevated concentrations of IL-1β, IL-6, tumor necrosis factor-α and decreased concentrations of glutathione


INTRODUCTION
Mastitis is a serious disease harming almost all mammals, including humans.In the dairy industry in particular, substantial losses in milk quality, animal health, and economics, as a result of very high treatment costs and culling rates, are caused by mastitis (Kim et al., 2019;Zhang et al., 2020).Previous studies indicated that infection of the mammary gland by exogenous pathogens was the major factor causing mastitis (Smith et al., 1985;Mungube et al., 2004).Thus, numerous studies and treatment measures for mastitis focused on the udder, including mammary gland tissue and milk (Eckersall et al., 2006;Addis et al., 2016;Martins et al., 2020).At present, the use of antibiotics has become the primary treatment on dairy farms, which has led to an increase in antibiotic residues and bacterial resistance in milk (Oliver and Murinda, 2012).However, it has gradually become clear that the external environment may not be the only source of bacteria in the udder.Studies on ruminants and mice have shown that certain gut microorganisms can enter the mammary gland via hematogenous and lymphatic translocation (Fox et al., 2005;Young et al., 2015;Hu et al., 2020).
The gastrointestinal tract of the host is colonized by large numbers of microorganisms.A crucial role of the gut microbiota is to modulate the host immune response (Kamada et al., 2013).Dysbiosis in the gut microbiota has been shown to increase the intestinal permeability and the level of systemic bacterial products, resulting in chronic inflammation (Frazier et al., 2011).Dysbiotic intestinal microbiota-induced inflammatory responses can be observed in multiple tissues and organs of the host, including serum, spleen, colon, and mammary gland (Ma et al., 2018).Several studies have reported the correlation among gastrointestinal microbiota and extra-intestinal diseases, such as multiple sclerosis (Lee et al., 2011), arthritis (Wu et al., 2010), and acute pancreatitis (Tsuji et al., 2012).These correlations are largely involved in the migration of immune cells (Kamada et al., 2013).Recently, the entero-mammary pathway was described and supported by several studies (Fernández et al., 2013;Jost et al., 2014;Young et al., 2015).Some bacteria from the gastrointestinal tract are capable of spreading to the mammary gland during lactation.This mainly relies on the mononuclear immune cells, especially dendritic cells, which can sample and carry several intestinal microbiota to the mammary gland (Rescigno et al., 2001;Perez et al., 2007).Microorganisms in different ecological niches of the host are demonstrated to be interconnected and constantly communicating (Costello et al., 2009).A study found that bacteria belonging to the Ruminococcus, Bifidobacterium, and Peptostreptococcaceae taxa all existed in milk, blood, and feces from the same cow (Young et al., 2015).These bacteria are usually considered part of the gut microbiota (Young et al., 2015).Similarly, Bifidobacterium have been observed in both milk and feces in humans (Martín et al., 2003;Perez et al., 2007).
To examine the gut inner environment (microbiota and metabolites) and inflammatory reaction during mastitis, we collected feces and blood samples from healthy cows and those with subclinical and clinical mastitis.Then, we used 16S rRNA gene sequencing and untargeted metabolomics technology to analyze the profile of fecal microbiota and metabolites in both feces and serum.These data provide further understanding of the gut microbial community and metabolite profiles of dairy cows with differing udder health status, as well as a new perspective for mastitis treatment.

Ethics Statement
The experimental design and operations were approved by the Animal Ethics Committee of the Chinese Academy of Agricultural Sciences (Beijing, China; approval number: IAS-2021-8).Sample collection was in accordance with the recommendations of the academy's guidelines for animal research.

Experimental Animals and Design
The present study was conducted on a Sino-Israel demonstration dairy farm (Beijing, China) during August 2021.Holstein dairy cows were milked 3 times a day at 0700, 1330, and 1900 h with an automatic milking system (Afimilk).Milk yield was recorded and milk samples collected for 7 consecutive days to analyze milk SCC, differential SCC (DSCC), and California Mastitis Test (CMT).Mastitis was diagnosed according to milk SCC, DSCC, and CMT results, as well as clinical udder symptoms.Currently, the optimal SCC cutoff point to distinguish between infected and uninfected quarters is 200,000 cells/mL (Turk et al., 2012;Zecconi et al., 2019).In addition, DSCC is recognized as a potential auxiliary indicator of SCC to accurately describe the udder inflammatory status of dairy cows; the DSCC represents the combined proportion of PMN and lymphocytes expressed as a percentage (Damm et al., 2017;Zecconi et al., 2021).The assay for DSCC followed the method described by Damm et al. (2017).The criteria for the CMT method followed Leach et al. (2008).Cows with similar parity, DIM, BW, and other factors, except for udder health status, were divided into 3 groups: (1) healthy group [H, n = 15; milk SCC <200,000 cells/mL; DSCC = 43.7 ± 1.62 (mean ± SEM); with negative CMT results and without fever, redness, or swelling in the udder]; (2) subclinical mastitis group [SM; n = 15; milk SCC >200,000 cells/mL; DSCC = 62.0 ± 1.91 (mean ± SEM); with positive CMT results and without fever, redness, or swelling in the udder]; and (3) clinical mastitis group [CM; n = 15; milk SCC >200,000 cells/mL; DSCC = 75.5 ± 1.14 (mean ± SEM); with strong positive CMT results and fever, redness, or swelling in the udder].Basic information on physical status of cows is listed in Supplemental Table S1 (https: / / data .mendeley.com/datasets/ pc79byc827/ 1; Wang, 2022).The proportions of PMN and lymphocytes in milk samples from H, SM, and CM groups are shown in Supplemental Figure S1 (https: / / data .mendeley.com/datasets/ pc79byc827/ 1; Wang, 2022).To prevent the spread of inflammation among dairy herds, cows designated as H, SM, and CM were housed in 3 separate cowsheds.None of the cows had received treatments involving antibiotics or other drugs.All selected cows were offered the same TMR 3 times a day at 0730, 1400, and 1930 h, with a roughageto-concentrate ratio of 60:40 (Supplemental Table S2; https: / / data .mendeley.com/datasets/ pc79byc827/ 1; Wang, 2022).Orts were weighed and recorded every day to calculate feed intake.

Blood and Feces Sample Collection
Two blood samples (5 mL each) from the tail vein of each cow were collected into 2 coagulation-promoting tubes.The blood samples were held for 45 min at ambient temperature (Dervishi et al., 2017).Sterile gloves were used to collect feces samples from the rectum of each cow, and the first few streams of feces were discarded before sample collection.Sampled feces were collected into sterile airtight bags, which were placed in a foam box with ice packs (Xu et al., 2017).The blood samples were centrifuged at 1,300 × g and 4°C for 15 min, and 3 aliquots of serum were obtained.Two were frozen at −20°C for assay of the concentration of inflammatory cytokines, lipids, oxidative stress indicators, and LPS; the third was stored at −80°C for analysis of the serum metabolome.Fecal samples from each cow were divided into two 5-mL sterile tubes, which were frozen at −80°C for analysis of fecal microbiota and metabolites.

16S rRNA Gene Sequencing of Fecal Microbiota
Fecal microbial DNA was extracted by using the BIOG DNA stool kit (Bioline) according to the kit instructions.Briefly, 200 mg of thawed fecal sample was added to a 5-mL sterile centrifuge tube with 2 mL of PBS and centrifuged at 300 × g at ambient temperature for 5 min.The supernatant was transferred to a 1.5-mL centrifugal tube, and centrifuged at 1,200 × g for 5 min.The pellet was resuspended with 1 mL of PBS, followed by centrifugation at 1,200 × g for 5 min.The pellet was transferred to a 3-mL centrifuge tube with 200 μL of lysis buffer (from the kit), and held at 37°C for 5 min.Subsequently, 600 μL of a solution containing 4.5 mL of NaCl and 25.5 mL of absolute ethyl alcohol was added and thoroughly mixed by vortex.An adsorption column containing 600 μL of the above solution was transferred to a collection tube, held for 2 min, and centrifuged at 1,200 × g and 4°C for 1 min.The remaining steps followed the manufacturer's protocol.Total DNA was eluted with 100 μL of Tris-EDTA buffer.A NanoDrop2000 spectrophotometer (Thermo Fisher Scientific) was used to determine DNA purity.The 16S V3-V4 variable region was selected for PCR amplification using primers 338F and 806R (338F: ACTCCTACGGGAGGCAGCAG; 806R: GGACTACHVGGGTWTCTAAT).Then, 2% agarose gel electrophoresis (Biowest), AxyPrep DNA gel extraction kit (Axygen), and Quantus Fluorometer were used to detect, purify, and quantify PCR products, respectively.Library construction was performed according to the instructions of NextFlex Rapid DNA-SEQ Kit (Bioo Scientific).

Untargeted Liquid Chromatography-MS Metabolomic Analysis
The thawed feces (50 mg) and serum (100 μL) samples were added to two 1.5-mL centrifuge tubes with 400 μL of methanol: water (4:1, vol/vol) solution to extract metabolites.The samples were mixed by vortex for 30 s and ultrasonication at 40 kHz for 30 min at 5°C.The samples were placed at −20°C for 30 min to precipitate the proteins.Then, 120 μL of reconstitution fluid (acetonitrile: water = 1:1, vol/vol) was added into the serum samples, followed by centrifugation at 13,000 × g and 4°C for 15 min.The supernatant was transferred to autosampler vials for liquid chromatography (LC)-MS analysis.All samples were mixed at equal volume to prepare quality control (QC) samples, which were inserted in every 3 samples.

Metabolomics Data Processing
The raw data were imported into Progenesis QI 2.3 software (Nonlinear Dynamics, Waters) for peak detection and alignment.The preprocessing results generated a data matrix that consisted of retention time, mass-to-charge ratio (m/z) values, and peak intensity.The response intensity of the sample mass spectrum peaks was normalized by sum to obtain a normalized data matrix.Metabolic features with a relative standard deviation of QC >30% were discarded.The information of metabolite features was identified using accurate mass, MS/MS fragment spectra, and isotope ratio difference by searching in reliable biochemical databases including the Human Metabolome Database (HMDB; http: / / www .hmdb.ca/ ) and Metlin database (https: / / metlin .scripps.edu/).

Statistical Analysis
Data of cows' health status, including DIM, BW, milk yield, and SCC, as well as concentrations of serum inflammatory cytokines, antioxidant indices, lipids, and LPS were analyzed using one-way ANOVA with Tukey's honestly significant difference as post hoc test procedure using SPSS software (version 22.0, IBM Corp.).Differences were declared significant at P < 0.05, whereas a tendency was assumed for 0.05 ≤ P < 0.10.
The Kruskal-Wallis H test with Tukey-Kramer post hoc multiple comparison analysis was used to assess significantly different microbiota among 3 groups.False discovery rate (FDR) was used to correct the P-value.Differences with FDR-adjusted P < 0.05 were considered significant.Linear discriminant analysis effect size (LEfSe) and linear discriminant analysis (LDA) were used to detect differences in microbial abundance between different groups and to determine the effect of the different microbiota on differences between groups (Wu et al., 2016).
For metabolomics data, a multivariate statistical analysis was performed using the ropls (version1.6.2, http: / / bioconductor .org/packages/ release/ bioc/ html/ ropls .html)R package from Bioconductor on the Majorbio Cloud Platform (https: / / cloud .majorbio.com).Multivariate statistical analyses, including principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA), were based on a unit variance scaling strategy, and with the 95% confidence level of Hotelling's T 2 statistic model (Sun et al., 2014).In PLS-DA plots, model validity was evaluated via 2 model parameters, R 2 and Q 2 , which provided information on the interpretability and predictability of the model, respectively; R 2 X(cum) and R 2 Y(cum) represent the cumulative interpretation rates for the X and Y matrices of the model, respectively.Significantly different metabolites between groups were defined based on fold change (>1.2 or <0.83), FDR-adjusted P-value (<0.05), and variable importance in projection value (VIP >1) calculated in the PLS-DA model.Metabolic pathway enrichment analysis was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (http: / / www .kegg.jp/kegg/ pathway .html).Spearman correlation analysis was performed using an online tool (https: / / cloud .majorbio.com/page/ tools/ ) to determine correlations (coefficients range from −1.0 to +1.0) between different fecal microbiota and metabolites in feces and serum, respectively.

Differences in Fecal Microbiota Among Cows in Different Groups
The difference of fecal microbial profiles among H, SM, and CM cows was initially distinguished using β-diversity analysis.As shown in Figure 2, PCoA and NMDS based on Bray-Curtis distance algorithm and analysis showed spatial separation in fecal microbiota among H, SM, and CM groups.The distances between the CM and H group samples were greater than those between the SM and H groups, indicating greater differences in fecal microorganism profiles between CM and H cows.

Metabolic Profile in Feces and Serum of Cows in Different Groups
Total ion chromatogram reflected the variability among biological repetitions and stability of the analyses in feces (Supplemental Figure S3A; https: / / data .mendeley.com/datasets/ pc79byc827/ 1; Wang, 2022) and serum (Supplemental Figure S3B) at the positive and negative ion modes.The overlap of retention time and total ion intensity of QC samples indicated the reliability of metabolomics data in the current study.Principal component analysis provided overall differences of metabolites profiles among H, SM, and CM groups.In PCA score plots of feces (Figure 4A) and serum samples (Figure 4B), symbols representing metabolites in the 3 groups were clearly distributed in different quadrants.The distance separating symbols between groups was much greater than that within groups, suggesting significant differences in metabolites in feces and serum among cows with different udder health status.To further identify differences in metabolites between groups, the data were processed by PLS-DA.The PLS-DA score plots in feces and serum samples show differences in the distribution of metabolites among H, SM, and CM groups (Supplemental Figure S4A and  C; https: / / data .mendeley.com/datasets/ pc79byc827/ 1; Wang, 2022).The model parameters R 2 X(cum) and R 2 Y(cum) represent the cumulative interpretation rate to the X and Y matrices of the model, respectively.The R 2 X(cum) and R 2 Y(cum) values were >0.5, indicating a high predictive ability.Response permutation testing was used to verify the accuracy of PLS-DA models.The value of Q 2 <0 in response permutation testing suggested that the PLS-DA model was not overfitted (Supplemental Figure S4B and D).

Metabolic Pathway Analysis of Different Metabolites
Metabolic pathways significantly enriched in different metabolites in feces and serum are listed in Tables 3 and 4, respectively.In feces and serum, metabolites involved in arachidonic acid metabolism, linoleic acid metabolism, and PBA biosynthesis pathways were up-regulated in CM and SM groups.In contrast, metabolites involved in purine metabolism, carnitine metabolism, citrate cycle (tricarboxylic acid cycle) and alanine, aspartate, and glutamate metabolism pathways were downregulated in both CM and SM groups.In addition, SBA biosynthesis pathways were downregulated in the CM group.

Serum Inflammatory Cytokines, Antioxidant Indices, Lipids, and LPS Concentrations
In the serum of CM cows, concentrations of IL-1β (P = 0.038), IL-2 (P = 0.031), TNF-α (P = 0.046), TC (P = 0.039), and TG (P = 0.026) were increased, and concentrations of IL-10 (P = 0.044), GSH-Px (P = 0.042), and HDL (P = 0.041) were decreased compared with cows in the H group. Concentrations of TNF-α in the CM and SM groups were not different (Table 5).The concentration of LPS in serum and feces of CM cows was increased compared with that in H (P = 0.007; P = 0.014) and SM cows (P = 0.041; P = 0.037), and serum LPS concentration in SM cows was elevated compared with that of H cows (P = 0.025; Figure 7A and B).

DISCUSSION
Previous studies reported the dysbiosis of milk microbiota during mastitis, which was considered to result from contamination by exogenous pathogenic bacteria (Zadoks et al., 2005;Zaror et al., 2011).Our observations further showed changes in some endogenous microorganisms from the host gut during mastitis, suggesting that the inflammatory response during mastitis may not be limited to the mammary gland.In the current study, Proteobacteria in feces was the main bacterial phylum that differed among H, SM, and CM cows.Recent studies have indicated that abnormal expansion of Proteobacteria promotes intestinal inflammation, which might relate to dysregulation of the immune response (Shin et al., 2015).Thus, increased Proteobacteria is regarded as a microbiological signal of intestinal dysbiosis (Shin et al., 2015).In the phylum Proteobacteria, 2 main opportunistic pathogens, Escherichia-Shigella and Klebsiella, were found to be significantly enriched in the feces of CM and SM groups, respectively; these are gram-negative bacilli that cause clinical mastitis in dairy cows (Pang et al., 2018).In addition, increased Paeniclostridium, classified as Clostridium sordellii, was believed to be a pathogenic bacterium in feces (Kim et al., 2017).In ruminants, Paeniclostridium was reported to be associated with intestinal inflammation-related diseases.Although the correlation between these gut microbiota and mastitis is not clear, the current findings showed significant enrichment of potentially pathogenic bacteria in the feces of mastitic cows, which might be involved in mastitis modulation.
Compared with healthy cows, lower community diversity and richness of gut microbiota were detected in SM and CM cows, which might be attributed to decreased feed intake in mastitic cows, especially in CM cows in the current study.Several studies have reported reduced activity, lying behavior, DMI, eating time, ruminating, and milk yield following mastitis (Bareille et al., 2003;Zimov et al., 2011;Yeiser et al., 2012).Lin et al. (2021) suggested that lesser dietary variety was significantly associated with reduced relative abundance of potentially beneficial bacteria and an increase in potentially harmful bacteria.Depleted fecal microorganisms were mainly from Firmicutes, Actinobacteriota, and Bacteroidota.Bacteria of the phylum Firmicutes, including Romboutsia, Lachnospiraceae_NK3A20_group, Lachnospiraceae_NK4A136_group, Coprococcus, and Eubacterium_ruminantium_group, are the main butyrate-producing bacteria (Duncan et al., 2002;Gasaly et al., 2021).Butyrate in the gut can inhibit production of proinflammatory cytokines by neutrophils, reduce neutrophil migration, and attenuate inflammation (Li et al., 2021), suggesting that a reduction in butyrate-  producing bacteria of the phylum Firmicutes might increase disease susceptibility.In addition, the lesser abundance of Lactobacillus detected in both SM and CM cows' feces was in line with findings by Ma et al. (2016), who reported that the lack of Lactobacillus is a common feature in milk and feces microbiota of mastitic cows.Another decreased bacterium in the feces of mastitic cows was Bifidobacterium, belonging to Actinobacteriota.Similar observations were reported in our previous study on milk from mastitic cows (Wang et al., 2020), suggesting some similarities in changes in feces and milk microbiomes.Bifidobacterium isolated from both milk and feces of the same host has been reported in other studies (Young et al., 2015).These findings might indicate the possibility of bacteria translocation from the gut to the mammary gland.Based on these findings, it was speculated that the decrease of 2 probiotics (Bifidobacterium and Lactobacillus) in the gut might be detrimental to host resistance to mastitis.Moreover, decreased Prevotellaceae_UCG-003 and Alistipes belonging to the phylum Bacteroidota are symbiotic bacteria in healthy intestines (Shkoporov et al., 2015;Accetto and Avguštin, 2019).The symbiotic microbiome protects against infection or inflammation in the immune response by inducing the anti-inflammatory cytokine IL-10 ( Mazmanian et al., 2008).
The current findings showed that potential pathogens increased and probiotic and symbiotic bacteria decreased during mastitis.The correlation between the gut microbiota and the occurrence and development of mastitis might be related to the entero-mammary gland pathway (Rodríguez, 2014).Bacterial translocation during lactation is considered a physiological event, especially in diseased individuals, where pathogenic bacteria can spread to the whole body (Lichtman, 2001).However, bacterial translocation can also occur in healthy individuals, where it is believed to be beneficial because it may be linked to immune regulation (Berg, 1995;Rodríguez et al., 2001).Based on previous studies and our observations, it is reasonable to believe that the gut microbiota changes during mastitis.
Inflammation and infection are usually accompanied by oxidative stress (Atroshi et al., 1996;Sordillo and Aitken, 2009) and changes in lipid metabolism (Memon et al., 2000).We found lower concentrations of GSH-P and SOD in the serum of mastitic cows.During mastitis, the increase in oxidative stress levels may be related to the excessive production of reactive oxygen species by neutrophils in the process of inflammation (Sordillo and Aitken, 2009).Decreased SOD and GSH-Px might contribute to oxidative damage (El-Deeb, 2013).High levels of serum TG and TC and a low level of HDL were observed in mastitic cows in the current study, which has also been observed in previous studies on mastitis     ( El-Deeb, 2013;Xiao et al., 2017).Variation in serum lipids during mastitis may involve the following 2 factors.First, as a typical inflammation mutagen, LPS can rapidly increase serum TC levels by stimulating liver production of very low density lipoprotein and reducing TC clearance (Xiao et al., 2012;Memon et al., 2000).In the present study, increased LPS concentrations in were observed in serum and feces of mastitic cows.Vojinovic et al. ( 2019) reported a correlation between the levels of circulating TC and HDL and gut microbiota.They found that serum HDL was positively correlated with genera from Lachnospiraceae and negatively correlated with Ruminococcus.Serum TG was also positively cor-related with Ruminococcus and negatively correlated with Coprococcus (Vojinovic et al., 2019).Ruminococcus is associated with the low abundance of gut microbes (Le Chatelier et al., 2013).Lachnospiraceae is one of the major taxa of intestinal microbes, participating in the maintenance of intestinal health and the conversion of PBA to SBA (Jia et al., 2018).The possible mechanism by which the gut microbiota affects circulating lipid levels involves modulation of bacteria-derived bile acids absorbed into blood (Allayee and Hazen, 2015;Fu et al., 2015;Ghazalpour et al., 2016).Similarly, during mastitis, changes in bile acid levels in feces and milk were observed in our study.Means within a row with different letters differed significantly (P < 0.05).Metabolites in the feces and serum of mastitic cows displayed similarities and differences.Elevated PBA and decreased SBA were observed in feces and serum of SM and CM cows.We found several positive correlations between certain fecal microbiota and SBA.The gut microbes involved in the conversion of PBA to SBA include Bifidobacterium, Lactobacillus, Bacteroides, and several bacteria from the Lachnospiraceae and Ruminococcaceae families (Gérard, 2013;Jia et al., 2018;Sinha et al., 2020).Secondary bile acids have wide range of anti-inflammatory effects and can suppress the expression of key cytokines and chemokines involved in inflammation, which is mainly related to bile acid signal receptors, such as farnesoid X receptor (FXR) and Takeda G protein-coupled receptor 5 (TGR5) (Hylemon et al., 2009;Sinha et al., 2020).Loss of SBA is thought to lead to inflammatory cytokine activation (Hagan et al., 2019).Thus, it is reasonable to speculate that lower levels of SBA metabolites in mastitic cows were related to the decrease in Bifidobacterium, Lactobacillus, Bacteroides, and Lachnospiraceae.
More metabolites involved in arachidonic acid and linoleic acid metabolism were enriched in feces and serum of mastitic cows compared with healthy cows.In the correlation analysis, the potential mastitiscausing bacteria Escherichia-Shigella, Streptococcus, Paeniclostridium, Klebsiella, and Corynebacterium showed positive associations with these proinflammatory metabolites.Among them, 20-trihydroxyleukotriene-B4 and 12-oxo-20-dihydroxy-leukotriene B4 were observed in both feces and serum.These are metabolites of lipid oxidation of leukotriene B4 (LTB4), which is the major metabolite of neutrophil polymorphonuclear leukocytes (Toda et al., 2002).In feces, 13,14-hihydro-15-keto-PGE 2 is a prostaglandin E 2 metabolite, which elicits diverse actions including pyrexia, pain sensation, and inflammation (Kawahara et al., 2015).Moreover, 9,10-DiHODE, 13-HpODE, and 13(S)-HODE, involved in linoleic acid metabolism, are proinflammatory lipid oxidation products that participate in TNF-α and IL-1 signaling pathways and mediate inflammation (Friedrichs et al., 1999).In the current study, higher concentrations of TNF-α and IL-1β were detected in the serum of CM cows compared with H cows. Compared with SM cows, more oligopeptides (Ac-Ser-Asp-Lys-Pro-OH, glycyl-histidine, isoleucyl-tyrosine) were found in the gut of CM cows.High levels of oligopeptides and free AA in the gut are reported to be associated with the activity of endogenous or bacterial-derived proteolytic enzymes (Moussaoui et al., 2002).A similar phenomenon has been observed in the milk of mastitic cows (Larsen et al., 2010).
In serum, higher levels of metabolites involved in sphingolipid and glycerophospholipid metabolism were found in CM cows, most of which were potent lipid proinflammatory mediators.Lysophospholipids are involved in leukocyte differentiation and activation, and they stimulate the expression of IL-1β and TNF-α in macrophages (Lee et al., 2002).In addition, S1P induces IL-6 secretion and cyclooxygenase-2 (COX-2) expression and regulates the production of arachidonic acid proinflammatory mediators (Völzke et al., 2014;Hsu et al., 2015).Similarly, lactosylceramide (d18: 1/ 18: 1(9Z)) is produced by lactate ceramide synthase activated by proinflammatory factors, and it participates in neutrophil migration and phagocytosis (Iwabuchi, 2018).The upregulation of these lipid proinflammatory metabolites in the serum of CM cows indicated a strong inflammatory process during mastitis.In the present study, increased levels of serum IL-1β, TNF-α, and IL-6 in mastitic cows might be attributed to high levels of proinflammatory mediators.
In the current study, downregulated metabolites in the feces and serum of mastitic cows mainly participated in the citrate cycle and carnitine and purine metabolism.Citric acid and carnitine are involved in the tricarboxylic acid cycle and β-oxidation of long-chain fatty acids, respectively (Akram, 2014;Longo et al., 2016), suggesting a retardation of energy metabolism during mastitis.Several purine metabolites, including inosine, are capable of inhibiting multiorgan inflammation (da Rocha Lapa et al., 2012).
In the current study, we observed changes in the gut microbiota during mastitis.Significantly different metabolites in the gut and serum of mastitic cows displayed some similarities, including elevated lipid proinflammatory products and decreased SBA.These data suggest that the inflammatory reaction during mastitis might not be limited to the mammary gland but may be systemic, including the gut and blood.

CONCLUSIONS
Fecal microbiota profiles diverged among H, SM, and CM cows in the present study.Pathogens and potential proinflammatory bacteria (Paeniclostridium, Escherichia-Shigella, and Klebsiella) were enriched in the gut of mastitic cows, whereas fewer probiotics and beneficial symbiotic bacteria (Bifidobacterium,Romboutsia,Lachnospiraceae_NK3A20_group,Coprococcus,and Alistipes) accounted for the decreased community diversity of gut microbiota.High levels of lipid proinflammatory metabolites and lower levels of SBA and compounds involved in energy and purine metabolism were found in both feces and serum of mastitic cows.In addition, decreased SOD, GSH-Px, and HDL and increased TC and TG in serum reflected the oxidative stress and lipid metabolism changes that accompany mastitis.These data contribute to a better understanding of the mechanism of mastitis in lactating dairy cows.
Wang et al.: FECAL MICROBIOME, METABOLOME, AND MASTITIS Figure 1.Community richness and diversity of feces sample from healthy (H), subclinical mastitis (SM), and clinical mastitis (CM) cows at the operational taxonomic units (OTU) level (n = 15).Bacterial richness was estimated by ACE and Sobs (observed species richness) values (A and B).Bacterial diversity was estimated by Simpson and Shannon indexes (C and D).Boxplots show bacterial community alpha-diversity indexes.The horizontal bars through the boxes show the median (i.e., the 50th percentile) of the bacterial distribution.The lower and upper extents of the boxes indicate the 25th and 75th percentiles of the distribution, respectively.The upper and lower whiskers indicate the maximum and minimum values, respectively.Asterisks indicate significant difference between the 2 groups (*0.01 < P ≤ 0.05; **0.001 < P ≤ 0.01).

Figure 3 .
Figure 3.The composition of different microorganisms in feces samples among healthy (H), subclinical mastitis (SM), and clinical mastitis (CM) cows at (A) the phylum level and (B) the genus level (average relative abundance >0.10%) (n = 15).Asterisks indicate significant difference between the 2 groups.g_ = genus; p_ = phylum.*0.01 < corrected P ≤ 0.05.(C) Cladogram shows linear discriminant analysis effect size (LEfSe) analysis of different fecal microbiota among H, SM, and CM groups from phylum to genus level.(D) Linear discriminant analysis (LDA) score plot indicating the effects of different microbiota on difference among H, SM, and CM groups.

Figure 4 .
Figure 4. Principal component (PC) analysis of metabolites in (A) feces and (B) serum samples from healthy (H), subclinical mastitis (SM), and clinical mastitis (CM) cows (n = 15).R 2 X(cum) represents the cumulative difference explanation rate of the model for the variable.

Figure 5 .
Figure 5. Hierarchical clustering analysis for identification of different metabolites in (A) feces and (B) serum samples among healthy (H), subclinical mastitis (SM), and clinical mastitis (CM) cows.Each column in the figure represents an individual sample, each row represents a metabolite, and the color indicates the relative abundance of metabolites expressed in the group; red indicates that the metabolite is expressed at high levels, and blue indicates lower expression.

Figure 6 .
Figure 6.Correlation analysis between (A) different fecal microbiota and fecal metabolites, and (B) different fecal microbiota and serum metabolites.Red represents a positive correlation, and blue represents a negative correlation.*Significant correlation between the concentrate and forage groups (false discovery rate-adjusted P < 0.05).